Hepatobiliary dysfunction remains the most life-threatening complication of parenteral nutrition (PN) in children (1–4). Cholestasis is the most studied aspect of PN-associated liver disease (5,6). It is easily diagnosed using biologically specific markers before the late appearance of hyperbilirubinemia and jaundice. Numerous risk factors are involved, which include infections that originate from the intestine or the intravenous catheter. In infants especially, early and repeated infections are considered risk factors for PN-associated cholestasis (7,8). Chronic PN-associated liver disease can lead to fibrosis, cirrhosis, and liver failure. However, factors predisposing to the development of liver fibrosis are not completely understood (9–11). Therefore, the purpose of this study was to assess the role of bacterial infections in the development of PN-associated liver fibrosis.
PATIENTS AND METHODS
Thirty patients enrolled in long-term PN programs between March 1985 and March 2000 were included in this retrospective study. The criteria for inclusion were long-term PN (>6 months), total PN (ensuring >80% of the protein and energy needs), and at least 1 liver biopsy (LB) available. The more recent LB was considered as the reference LB. Its date marked the end of the survey for each child. Children with inborn liver diseases or acquired viral or toxic liver diseases were not included.
Reference LB was performed in 7 children at the time of isolated intestinal transplantation, in 5 children at the time of combined intestinal and liver transplantation before liver removal, and in the other 18 children at the time of the decision whether to perform isolated intestinal transplantation or combined intestinal and liver transplantation. At the time of the reference LB, the median age of the patients was 44 months (range, 8–152 months). Parenteral nutrition was started within the neonatal period in 29 children and after 5 months of life in 1 case (median, 1 month; range, 1–5 months of life). Twenty-four patients were full-term babies and 6 were premature infants (30–36 weeks of gestation). The indications for PN were intractable diarrhea (n = 12), short bowel syndrome (n = 8), long-segment Hirschsprung disease (n = 6), and chronic intestinal pseudoobstruction (n = 4; Table 1).
Median PN duration was 44 months (range, 8–150 months). Parenteral nutrition was delivered through central venous Broviac-type catheters, usually on 12-hour or 16/24-hour cycles nightly. All of the patients were included in home PN (HPN) programs. Before 1995 HPN consisted of binary mixtures (including glucose, amino acids, electrolytes, trace elements, and vitamins), whereas lipid emulsions were administered separately through a Y-catheter. From 1995 on all-in-one mixtures were provided to patients receiving HPN. Mean PN supplies were calculated from the data recorded between 12 and 18 months of life for each patient (Table 1). The energy-to-nitrogen ratio was 200:250 kcal/g nitrogen in both groups. The lipids were usually infused 3–5 days/week and provided 15% to 30% of nonprotein energy supply. The first-choice lipid emulsions were pure long-chain triglyceride emulsions. Lipid delivery was stopped each time toxicity was detected, such as cholestasis and/or lipid overload syndrome or an immune dysfunction related to medullar hyperactivation of the macrophages induced by intravenous lipid emulsions (12,13). In these cases, after normalization of liver blood test results and/or platelet count, lipid delivery was restarted with a mixed emulsion of 50% medium-chain triglyceride and 50% long-chain triglyceride 1–3 days/week.
The following medications were given when needed: ursodeoxycholic acid (n = 9), ranitidine (all cases), cholestyramine (n = 7), metronidazole and colimycin (n = 9), sulfamides as urinary antiseptics (4 cases of pseudoobstruction and 3 cases of urinary tract infections with ureteral reflux), somatostatin (n = 1), azathioprine (n = 2), cyclosporine (n = 2), gancyclovir (n = 1), acyclovir (n = 1), and corticosteroids (n = 4).
Sepsis episodes were defined as episodes of fever with clinical and biological parameters of infection (eg, increased plasma concentration of C-reactive protein and hyperleukocytosis), positive blood cultures from the catheter and peripheral vein for the same pathogen or >1 positive blood culture from the catheter for the same pathogen, infection at the exit site of the catheter with a positive site culture finding with or without positive blood cultures for the same organism, and positive response to antibiotics. Episodes that shared the same characteristics but without microorganisms found in blood culture samples and that showed clinical and biologic improvement under antibiotic treatment were considered to constitute sepsis as well. Infections were classified according to the isolated microorganisms (eg, Staphylococcus species and other cocci, bacilli, and fungi; Table 2). In the absence of obvious alternative sites of primary infection, antibiotic treatment was immediately started while the results of blood cultures were awaited. Antibiotic treatment was delivered through the catheter and included at least 1 agent against Staphylococcus species, which was vancomycin. In the present study antibiotics were adapted according to the results of the antibiogram and included β-lactamins, aminoglycosides, rifamycin, glycopeptides, fosfomycin, metronidazole, amphotericin B, or fluconazole.
All of the LB samples were percutaneous needle biopsy samples. Liver fibrosis was graded according to the extent of healing lesions. Patients were distributed into 2 groups on the basis of pathologic criteria. Group A included patients in whom the LB showed severe septal fibrosis involving >50% portal fields or cirrhosis. Group B included patients with normal liver architecture or mild to moderate fibrosis (involving <50% of portal fields).
Analysis criteria for each patient were duration of pregnancy, birth weight, age at PN onset, underlying disease, pathological liver biopsy analysis, age at the date of LB, PN duration until LB, PN features, associated treatments, digestive surgical or occlusive episodes, biological cholestatic episodes (conjugated hyperbilirubinemia level >20 μmol/L), and occurrence of lipid overload syndrome. Infection histories were recorded for each patient.
Results are expressed as means ± SEM except for the age of PN onset, the age of the reference biopsy, PN duration, and the age of the first infection, which are expressed in medians and ranges. Statistical comparisons were performed using the nonparametric Mann-Whitney test. Percentages were compared by Fisher exact test. P is considered significant if lower than 0.05.
Group A included 16 patients with severe liver fibrosis or cirrhosis and group B included 14 patients with normal LB or mild septal fibrosis. Both groups were equivalent in terms of gestational age, birth weight, primary disease, age and indications for PN onset, and mode and rhythm of PN delivery (Table 1). Group A included 2 premature infants (30 and 34 gestational weeks) and group B included 4 premature infants (34–36 gestational weeks) and 1 infant who was small for gestational age. Duration of PN at the time of the reference LB was inversely correlated to the extent of liver fibrosis: 30.5 months (range, 8–96 months) in group A versus 105 months (range, 37–150 months) in group B (P < 0.001). The associated treatments and conditions related to the risks of infection were similar in both groups.
Of 324 infectious episodes reported, a microorganism was found in blood culture in 223 cases (68%): Staphylococcus species were found in 45%, other cocci in 7%, bacilli in 40%, and fungi in 8%. The other 101 episodes (32%) were considered possible infections. The incidence of infections was significantly higher in group A (P < 0.001) than in group B: 3.2 ± 0.3/study-year versus 1.5 ± 0.2/study-year (Table 2). In both groups most infections were caused by Staphylococcus species and Gram-negative bacilli.
The first infection occurred significantly earlier in group A than in group B (P < 0.01): 1 month (range, 1–2 months) versus 4 months (range, 1–19 months), respectively (Table 2). The first infection occurred before the age of 1 month in 75% of the children in group A versus only 21% in group B (P < 0.01). In group A 100% of the first infections occurred before 2 months of age, compared with 35% in group B (P < 0.001). Gram-negative bacilli were responsible for 50% of the first infections in group A versus 14% in group B (P = 0.05).
The incidence of surgical episodes with digestive stasis or obstruction was identical in both groups, as were episodes of fat overload syndrome (Table 3). Incidences of clinical cholestatic episodes were not statistically different between groups (81% of the patients in group A vs 50% in group B). Neither the age of cholestasis onset nor the total duration of cholestasis differed significantly. Meanwhile, according to LB, cholestasis more often existed in group A (68%) than in group B (21%; P < 0.05). In group A 4 LBs showed severe cholestasis, 5 showed moderate cholestasis, 2 showed mild cholestasis, and 5 did not show any sign of cholestasis but were already cirrhotic. In group B 11 of the 14 LBs did not show any signs of cholestasis, 1 showed moderate cholestasis, and 2 showed mild cholestasis.
Once PN-associated liver disease is established, it can progress to cirrhosis and liver failure (2,9). Numerous causes of PN-associated liver disease have been proposed (1,3–6). Many nutritional risk factors have been suggested, including excess of energy supply, type and amount of lipids, essential fatty acid deficiency, excess of amino acid supply, choline deficiency, zinc deficiency, manganese or aluminium toxicity, and the absence of enteral feeding; however, in the present study the mode of PN delivery was identical in both groups. Moreover, the incidence of fat overload syndrome was the same in both groups. Among patient-related risk factors, underlying digestive diseases, number of abdominal surgical procedures, and incidence of episodes of intestinal obstruction were similar in both groups.
The age of PN onset is also usually considered to be a risk factor (7,8). Numerous studies in infants with PN-associated liver disease have suggested that immaturity of the hepatobiliary tract (5) was a risk factor for cholestasis. In our study the differences in the extension of liver fibrosis between study groups could not be explained by an early beginning of total PN or prematurity. Parenteral nutrition was started early (median age, 1 month; range, 1–5 months of age) in the whole population, which could be therefore considered to be particularly exposed to PN-related toxicity.
Sepsis is another recognized risk factor for PN-associated liver disease related to central venous catheters (14) and underlying digestive diseases (15,16). A study in neonates with intestinal resection and long-term PN showed that the mean age at the first infection was significantly younger in cholestatic patients whose condition progressed to liver failure (1 month) than in noncholestatic patients (6 months of age) (8). The role of infections in the development of cholestasis was suggested in other previous series (7). However, the relationship between infections and liver fibrosis has not been studied so far. Many authors have proposed a course or even a timetable of hepatic histological findings beginning with cholestasis and culminating in cirrhosis (9). The present study failed to show a clear relationship between the number and duration of clinical cholestatic episodes and the extent of liver fibrosis, although LB showed that cholestasis was significantly more frequent in group A than in group B.
The present study also suggests that infections are associated with the development of liver fibrosis in children with early and prolonged PN, which is independent of duration of PN, mode of PN delivery, or type of underlying disease.
For each patient, the study period included time at home as well as hospital stay. This explains the higher rate of infections than reported in a study in patients receiving HPN in the same center, at approximately 2/1000 HPN-days (14).
In the group with severe fibrosis, the incidence of infections was twice as high as in the group with preserved liver function, which is inversely correlated with PN duration, suggesting a relation between the number of infections and liver fibrosis in the children with early and prolonged PN independent of other risk factors.
Among the microorganisms responsible for infections listed in this study, a majority were Staphylococcus species and Gram-negative bacilli. The latter were involved in 50% of the first sepsis episodes in group A. Recent studies in surgical newborns suggested an intestinal origin of septicemia (16,17). The carriage of abnormal flora is a risk factor for septicemia in the neonatal period (16). A 42% incidence of carriage of Gram-negative bacilli in neonates has been reported (16). Moreover, as many as 84% of the Gram-negative bacilli carrier newborns presented with septicemia (17). The present study suggests that septicemia before 6 weeks of life is a risk factor for further liver disease if PN is prolonged. Therefore, abnormal flora carrier infants may be particularly exposed to liver disease if they need long-term PN. If the first TPN-related case of septicemia is mainly a gut-derived phenomenon, then novel strategies for prevention are required. Small-volume enteral feeding (18), as well as specific nutrients, growth factors, or hormones (19,20), may be useful tools for the modulation of gut immunity. Modulation of the intestinal flora by selective digestive decontamination (21,22) or by prebiotics or probiotics is also a matter of debate.
In the whole population studied, a high rate of Staphylococcus infections was confirmed. In numerous studies this major cause of late sepsis is related to central venous catheters in long-term PN (23). By modulating immune function, PN may predispose children to Staphylococcus infections. Host defense mechanisms including phagocytosis and killing of coagulase-negative staphylococci have been shown to be impaired during long-term PN (24). Bacterial infections and exposure of the liver to endotoxins initiate a cascade of inflammatory events (25). The inflammatory response recruits mononuclear and hepatic stellar cells and contributes to tissue remodeling (26). If injury persists, then this activated process may lead to tissue scarring and development of liver fibrosis (27,28).
The present study suggests that in infants and children receiving long-term PN, early sepsis (<4 weeks of life) contributes to the development of severe hepatic fibrosis or even cirrhosis before the end of the third year of life. An initial Gram-negative Bacillus infection seems to be a particular risk factor. This suggests a progressive process initiated by the first infection and continued with further infections inducing fibrosis or even cirrhosis in children receiving long term-PN.
Our data should prompt pediatricians to prevent, diagnose, and aggressively treat all forms of infections, especially in infants who are thought to depend further on long-term PN. Moreover, liver function must be carefully monitored in children with a history of early and repeated sepsis.
1. Kelly DA. Intestinal failure-associated liver disease: what do we know today: Gastroenterology 2006; 130(Suppl 1):S70–S77.
2. Colomb V, Goulet O, Rambaud C, et al
. Long term-parenteral nutrition (PN) in children: liver and gallbladder disease. Transplant Proc 1992; 24:1054–1055.
3. Angelico M, Della Guardia P. Review article: hepatobiliary complications associated with total parenteral nutrition. Aliment Pharmacol Ther 2000; 14:54–57.
4. Shulman RJ. New developments in total parenteral nutrition for children. Curr Gastroenterol Rep 2000; 2:253–258.
5. Owings E, Georgeson K. Management of cholestasis in infants with very low birth weight. Semin Pediatr Surg 2000; 9(2):96–102.
6. Forchielli ML, Walker WA. Nutritional factors contributing to the development of cholestasis during total parenteral nutrition. Adv Pediatr 2003; 50:245–267.
7. Beath SV, Davies P, Papadopoulo A, et al
. Parenteral nutrition-related cholestasis in postsurgical neonates: multivariate analysis of risk factors. J Pediatr Surg 1996; 31:604–606.
8. Sondheimer JM, Asturias E, Cadnapaphornchai M. Infection and cholestasis in neonates with intestinal resection and long-term parenteral nutrition. J Pediatr Gastroenterol Nutr 1998; 27:131–137.
9. Mullick FG, Moran CA, Ishak KG. Total parenteral nutrition: a histological analysis of the liver changes in 20 children. Mod Pathol 1994; 7:190–194.
10. Meehan JJ, Georgeson KE. Prevention of liver failure in parenteral nutrition-dependent children with short bowel syndrome. J Pediatr Surg 1997; 32:473–475.
11. Kaufman SS. Prevention of parenteral nutrition-associated liver disease in children. Pediatr Transplant 2002; 6:37–42.
12. Goulet O, Girot R, Maier-Redelsperger M, et al
. Hematologic disorders following prolonged use of intravenous fat emulsions in children. JPEN J Parenter Enteral Nutr 1986; 10:284–288.
13. Colomb V, Jobert-Giraud A, Lacaille F, et al
. Role of lipid emulsions in cholestasis associated with long-term parenteral nutrition in children. JPEN J Parenter Enteral Nutr 2000; 24:345–350.
14. Colomb V, Fabeiro M, Dabbas M, et al
. Central venous catheter-related infections in children on long-term home parenteral nutrition: incidence and risk factors. Clin Nutr 2000; 19:355–359.
15. Pierro A, van Saene H, Jones M, et al
. Clinical impact of abnormal gut flora in infants receiving parenteral nutrition. Ann Surg 1998; 227:547–552.
16. van Saene HKF, Taylor N, Donnell SC, et al
. Gut overgrowth with abnormal flora: the missing link in parenteral nutrition-related sepsis in surgical neonates. Eur J Clin Nutr 2003; 57:548–553.
17. Donnell SC, Taylor N, van Saene HKF, et al
. Infection rates in surgical neonates and infants receiving parenteral nutrition: a five-year prospective study. J Hosp Infect 2002; 52:273–280.
18. Okada Y, Klein N, van Saene H, et al
. Small volumes of enteral feedings normalise immune function in infants receiving parenteral nutrition. J Pediatr Surg 1998; 33:16–19.
19. McAndrew HF, Lloyd DA, Rintala R, et al
. Intravenous glutamine or short-chain fatty acids reduce central venous catheter infection in a model of total parenteral nutrition. J Pediatr Surg 1999; 34:281–285.
20. McAndrew HF, Lloyd DA, Rintala R, et al
. The effects of intravenous epidermal growth factor on bacterial translocation and central venous catheter infection in the rat total parenteral nutrition model. Pediatr Surg Int 2000; 16:169–173.
21. Oudemans-van Straeten HM, van Saene HKF, Zandstra DF. Selective decontamination of the digestive tract: use of the correct antibiotics is crucial. Crit Care Med 2002; 30:38–43.
22. Donnell SC, Taylor N, van Saene HKF. Translocation cannot be ignored during parenteral nutrition. J Hosp Infect 2003; 54:165–167.
23. Nataro JP, Corcoran L, Zirin S, et al
. Prospective analysis of coagulase-negative staphylococcal infection in hospitalized infants. J Pediatr 1994; 125:798–804.
24. Okada Y, Klein NJ, van Saene HKF, et al
. Bactericidal activity against coagulase-negative staphylococci is impaired in infants receiving long-term parenteral nutrition. Ann Surg 2000; 231:276–281.
25. Romeo C, Eaton S, Quant PA, et al
. Neonatal oxidative liver metabolism: effects of hydrogen peroxide, a putative mediator of septic damage. J Pediatr Surg 1999; 34:1107–1111.
26. Williams EJ, Gaca MDA, Brigstock DR, et al
. Increased expression of connective tissue growth factor in fibrotic human liver and in activated hepatic stellate cells. J Hepatol 2000; 32:754–761.
27. Poli G. Pathogenesis of liver fibrosis: role of oxidative stress. Mol Aspects Med 2000; 21:49–98.
28. Wu J, Zern MA. Hepatic stellate cells: a target for the treatment of liver fibrosis. J Gastroenterol 2000; 35:665–672.