Bilirubin entry into and clearance from rat brain during hypercarbia and hyperosmolality

Molecular events in brain bilirubin toxicity revisited

The mechanisms involved in bilirubin neurotoxicity are still far from being fully elucidated. Several different events concur to damage mainly the neurons among which inflammation and alteration of the redox state play a major role. An imbalance of cellular calcium homeostasis has been recently described to be associated with toxic concentrations of bilirubin, and this disequilibrium may in turn elicit an inflammatory reaction. The different and age-dependent sensitivity to bilirubin damage must also be considered in describing the dramatic clinical picture of bilirubin-induced neurological damage (BIND) formerly known as kernicterus spectrum disorder (KSD). This review aims to critically address what is known and what is not in the molecular events of bilirubin neurotoxicity to provide hints for a better diagnosis and more successful treatments. Part of these concepts have been presented at the 38th Annual Audrey K. Brown Kernicterus Symposium of Pediatric American Society, Washington DC, May 1, 2023.

Models of bilirubin neurological damage: lessons learned and new challenges

OBJECTIVE

Jaundice (icterus) is the visible manifestation of the accumulation of bilirubin in the tissue and is indicative of potential toxicity to the brain. Since its very first description more than 2000 years ago, many efforts have been undertaken to understand the molecular determinants of bilirubin toxicity to neuronal cells to reduce the risk of neurological sequelae through the use of available chemicals and in vitro, ex vivo, in vivo, and clinical models. Although several studies have been performed, important questions remain unanswered, such as the reasons for regional sensitivity and the interplay with brain development. The number of new molecular effects identified has increased further, which has added even more complexity to the understanding of the condition. As new research challenges emerged, so does the need to establish solid models of prematurity.

METHODS

This review critically summarizes the key mechanisms of severe neonatal hyperbilirubinemia and the use of the available models and technologies for translational research.

IMPACT

We critically review the conceptual dogmas and models used for studying bilirubin-induced neurotoxicity. We point out the pitfalls and translational gaps, and suggest new clinical research challenges. We hope to inform researchers on the pro and cons of the models used, and to help direct their experimental focus in a most translational research.

Molecular Physiology and Pathophysiology of Bilirubin Handling by the Blood, Liver, Intestine, and Brain in the Newborn

Bilirubin is the end product of heme catabolism formed during a process that involves oxidation-reduction reactions and conserves iron body stores. Unconjugated hyperbilirubinemia is common in newborn infants, but rare later in life. The basic physiology of bilirubin metabolism, such as production, transport, and excretion, has been well described. However, in the neonate, numerous variables related to nutrition, ethnicity, and genetic variants at several metabolic steps may be superimposed on the normal physiological hyperbilirubinemia that occurs in the first week of life and results in bilirubin levels that may be toxic to the brain. Bilirubin exists in several isomeric forms that differ in their polarities and is considered a physiologically important antioxidant. Here we review the chemistry of the bilirubin molecule and its metabolism in the body with a particular focus on the processes that impact the newborn infant, and how differences relative to older children and adults contribute to the risk of developing both acute and long-term neurological sequelae in the newborn infant. The final section deals with the interplay between the brain and bilirubin and its entry, clearance, and accumulation. We conclude with a discussion of the current state of knowledge regarding the mechanism(s) of bilirubin neurotoxicity.

Prevention of neurodevelopmental sequelae of jaundice in the newborn

Although its cause, jaundice in the newborn, is extremely common, the disabling neurological disorder kernicterus is very rare. Kernicterus may be prevented by selecting those infants who are at risk of extreme jaundice or who may be particularly vulnerable to bilirubin neurotoxicity. Because the tools for achieving that goal are inadequate, a secondary strategy is needed. This involves a plan for emergency treatment of severely jaundiced infants, in particular those who present with neurological symptoms. In this paper I review the strategies for preventing extreme jaundice, and for reversing neurotoxicity in those infants for whom the principal strategies fail. Briefly, the tools for prevention include measurement of bilirubin while the infant is staying in the maternity unit, plotting the value on an hour-specific chart, assessing other risk factors for jaundice, and educating the parents. Emergency treatment should include immediate, high-irradiance phototherapy, consideration of intravenous immune globulin, and preparation for an exchange transfusion.

Calculated free bilirubin levels and neurotoxicity

Although most bilirubin in the circulation is bound to albumin, a relatively small fraction remains unbound. The concentration of this 'free' bilirubin (B(F)) is believed to dictate the biologic effects of bilirubin in jaundiced newborns, including its neurotoxicity. The threshold at which B(F) produces changes in cellular function culminating in permanent cell injury and cell death has been the subject of considerable debate. The objective of this study was to compare calculated central nervous system (CNS) B(F) levels in Gunn rat pups during (i) peak postnatal hyperbilirubinemia and (ii) sulfadimethoxine-induced acute bilirubin encephalopathy (ABE) previously reported from our laboratory with those predicted in human neonates with peak total serum bilirubin (TSB) levels of 35 mg per 100 ml (599 micromol l(-1)), a clinical cohort that often evidence moderate-to-severe adverse post-icteric neurodevelopmental sequelae. Homozygous j/j Gunn rat pups with neonatal hyperbilirubinemia due to a deficiency of the bilirubin conjugating enzyme uridine-diphosphate-glucuronosyl transferase 1A1 were studied along with non-jaundiced littermate heterozygous J/j controls. Sulfadimethoxine was used to displace bilirubin from albumin in hyperbilirubinemic j/j Gunn rat pups to increase their brain bilirubin content and induce ABE. Calculated Gunn rat CNS B(F) levels were determined as a function of genotype, sulfadimethoxine exposure and albumin-bilirubin binding constant. These data were compared with the human CNS B(F) predicted from the calculated serum B(F) in human neonates with a TSB of 35 mg per 100 ml as a function of albumin-bilirubin binding constant, albumin concentration and the assumption that at this hazardous bilirubin level there may be rapid equilibration of B(F) between serum and brain. There was a large gap between the upper limit of the calculated CNS B(F) 95% confidence interval (CI) range in non-jaundiced J/j pups (for example, 112 nM at k=9.2 l micromol(-1)) and the lower limit seen in the saline-treated hyperbilirubinemic j/j pups (556 nM at k=9.2 l micromol(-1)) as well as between the upper limit in saline-treated hyperbilirubinemic j/j pups (1110 nM at k=9.2 l micromol(-1)) and the lower limit seen in sulfadimethoxine-treated jaundiced j/j littermates (3461 nM at k=9.2 l micromol(-1)). There was considerable overlap and remarkable similarity between the predicted human CNS B(F) values at a TSB of 35 mg per 100 ml for a range of reported human serum bilirubin-albumin binding constants and albumin concentrations, and those calculated for saline-treated hyperbilirubinemic j/j Gunn rat pups. This exercise yielded strikingly similar apparent calculated neurotoxic B(F) levels for Gunn rat pups and human neonates rather than orders of magnitude differences that might have been predicted at the outset and add to a growing literature aimed at defining clinically germane neurotoxic B(F) thresholds.Journal of Perinatology (2009) 29, S14-S19; doi:10.1038/jp.2008.218.

Bilirubin-induced cell death during continuous and intermittent phototherapy and in the dark

AIM

To compare continuous and intermittent light exposure in the presence of bilirubin with respect to cellular damage. Furthermore, it was of interest to characterize the nature of cellular toxicity of bilirubin in the dark.

METHOD

A murine lymphoma cell line, L5178Y-R (LY-R), was exposed to solutions of bilirubin (160 microM) supplemented with human serum albumin (200 microM) and irradiated with phototherapy light (Philips 20W/52) at a constant total dose of approximately 500 kJ/m2. The irradiation was given either as intermittent or continuous treatment with light of variable irradiance. The three lower irradiance levels were clinically relevant. Cells treated with bilirubin were also kept in the dark for various periods of time. Cell toxicity was determined by measuring apoptosis and necrosis. Apoptosis was measured by terminal deoxynucleotide transferase and propidium iodide staining assay, while trypan blue assay was used for detection of necrosis.

RESULTS

There was no difference (n = 6, p > 0.05) between continuous and intermittent irradiation in the induction of early and late apoptotic cell death. Necrosis was more pronounced after intermittent treatment. Bilirubin dark toxicity was observed and classified as both apoptotic and necrotic.

CONCLUSION

Continuous and intermittent light exposure caused the same degree of apoptotic cell death, while the cells underwent more necrotic death after intermittent exposure. Bilirubin was cytotoxic in the dark by both cell death mechanisms.

Administration of drugs known to inhibit P-glycoprotein increases brain bilirubin and alters the regional distribution of bilirubin in rat brain

P-glycoprotein (P-gp) is an ATP-dependent integral plasma membrane efflux pump, expressed in abundance in brain capillary endothelial cells and astrocytes. P-gp contributes to the blood-brain barrier in limiting the influx and retention of a variety of lipophilic compounds, including unconjugated bilirubin. Several drugs block P-gp function and thereby enhance intracellular accumulation of P-gp substrates. In this study we proposed that pretreatment with drugs known to inhibit P-gp function in clinically relevant doses would alter the uptake of bilirubin in the brain of 32- to 36-d-old rats. In the first arm of the study, the animals received pretreatment with an i.v. infusion of either propanolol, erythromycin, verapamil, ceftriaxone, rifampin, or saline, 10 min before an i.v. bolus of 50 mg/kg bilirubin was given. Except for the erythromycin-treated rats, all treatment groups had significantly higher brain-to-serum bilirubin ratios than control animals (p < 0.05, Welch's t test). In the second arm of the study, treatment with either ceftriaxone or rifampin or saline i.v. preceded a 50 mg/kg i.v. bolus of radioactive bilirubin. Analysis of seven different brain regions by scintillation counting showed that the distribution patterns differed significantly between the study groups (p < 0.001, ANOVA), however, not in accordance with a kernicteric staining pattern. Because of limited knowledge of expression and function of P-gp and other membrane transport proteins in the newborn, the implications of this study remain to be seen. We speculate that drugs known to inhibit P-gp function may increase the risk of bilirubin encephalopathy in the hyperbilirubinemic infant.

The effects of bilirubin on evoked potentials and long-term potentiation in rat hippocampus in vivo

Neonatal jaundice is a common condition that could potentially lead to severe neurotoxicity. In this electrophysiological study we observed the effects of a short-term bilirubin injection on evoked potentials (population spike, PS) and long-term potentiation (LTP) in the hippocampal CA3 region of Sprague Dawley rats in vivo. The animal received a bolus i.v. injection of either 60 mg/kg, or 30 mg/kg of bilirubin, or an equivalent volume of bilirubin-free vehicle in 5 min. The results showed that both bilirubin-treated groups had a dose-independent prolongation of peak latencies and decrease of slopes of the PS at all measured time points following injection (1, 3, 5, 10, 15, 30, 45, 60, 90, 120 min), while the amplitudes of the PS did not change significantly. The peak latency, slope, and the amplitude of PS stayed unchanged in the control group. Furthermore, while LTP could be induced by high-frequency stimulation in control animals, this phenomenon was absent in both bilirubin-treated groups. The amplitudes of the PS in the two treated groups after stimulation were also smaller than those of the control animals at every time points. These findings are in accordance with previous observations showing significant depressive effects of bilirubin on the nervous system. Our novel finding that short-term exposure to bilirubin can inhibit the induction of LTPs in the hippocampus, is compatible with the suggestion that neonatal hyperbilirubinemia can impact on learning and memory.

Mechanisms of bilirubin toxicity: clinical implications

The basic mechanism of kernicterus and bilirubin encephalopathy has not been unequivocally determined. Much knowledge has been gained about phenomena that contribute to bilirubin neurotoxicity, and this knowledge has implications for clinical practice. Conditions that impact on blood-brain barrier function, increase brain blood flow, or impact on bilirubin metabolism, including its transport in serum, should be avoided, if possible. Such conditions include drugs and drug stabilizers that compete with bilirubin binding to albumin, or that inhibit P-glycoprotein in the blood-brain barrier, prematurity/immaturity, and clinically significant illness in the infant that involves hemolysis, respiratory and metabolic acidosis, infection, asphyxia, hypoxia and (perhaps) hyperoxia, and hyperosmolality. If these conditions are not avoidable then there should be a more aggressive approach to the treatment of hyperbilirubinemia. The limits of tolerance for hyperbilirubinemia varies among neonates and there are no tools to determine with certainty when a particular infant is approaching the danger zone. Neurological symptoms in a jaundiced infant require extreme vigilance, and, in most cases, immediate intervention.

Kernicterus: an international perspective

Kernicterus occurs in all parts of the world. The risk is increased in countries where glucose-6-phosphate dehydrogenase-deficiency is common. In the 1990's more case reports of infants who developed kernicterus were published than in the previous decades. A combination of reduced concern for jaundice in newborns, early discharge with inadequate follow-up and a decreased awareness of the signs that may herald serious toxicity may have contributed to the apparent increase in the incidence of kernicterus. Although most jaundiced newborns do not need aggressive evaluation or treatment, physicians who deal with such infants still need to be vigilant. We lack the necessary tools to distinguish infants who may be particularly vulnerable to the effects of bilirubin on the brain from those who may tolerate high serum bilirubin levels without harm. Therefore it is imperative that each infant with significant jaundice be conscientiously evaluated and a plan for testing and, if necessary, therapy be formulated. Transcutaneous measurement of bilirubin is a simple tool that significantly reduces the need for invasive tests. Signs of possible neurotoxicity must never be disregarded or neglected. Any jaundiced infant who shows signs of possible neurotoxicity should receive intensive phototherapy as an emergency procedure while further evaluation continues. Further studies regarding bilirubin toxicity and neonatal jaundice are needed, both in the basic science as well as in the clinical arena.

Bilirubin brain toxicity

Bilirubin is toxic in most biological systems tested. Several mechanisms have been suggested for this toxic effect, including inhibition of enzyme systems and inhibition of cell regulatory reactions (protein/peptide phosphorylation). The identity of the basic mechanism(s) has not been conclusively proven, but inhibition of peptide phosphorylation, perhaps mediated or modulated by lysine at the active site(s), appears to be compatible with many of the observations currently found in the literature. Bilirubin entry into brain is facilitated by drug displacement of bilirubin from its albumin binding, reduced albumin binding capacity, increased brain bloodflow, increased permeability of the blood-brain barrier, and other factors. The rate of bilirubin entry into brain, as well as the degree of retention and rate of clearance from brain, depends on which of these circumstances are operative. It is as yet unclear whether the mechanism responsible for increased brain bilirubin is important for toxicity. The mechanism for preferential localization of bilirubin to the basal ganglia in kernicterus is also not known. Bilirubin appears to distribute differentially to brain subcellular compartments and is oxidized in brain by an enzyme localized on the inner mitochondrial membrane. This enzyme is found both in neurons and in glia, but appears to be more active in the latter. The activity increases with postnatal age, and is subject to genetic variability in animals. The enzyme is cytochrome c-dependent. It is as yet not clear whether the activity of this enzyme serves a brain-protective effect in severe hyperbilirubinemia.

Subcellular localization of bilirubin in rat brain after in vivo i.v. administration of [3H]bilirubin

Bilirubin appears to be toxic in vitro to several cellular functions localized to different subcellular compartments. It would therefore be useful to know what concentrations of bilirubin might be found in cell organelles in vivo. Rats were anesthetized and allocated to one of three groups: control, hypercarbia, and hyperosmolality. Each rat received a 5-min bolus dose of bilirubin 50 mg/kg i.v. (containing approximately 200 microCi [(3)H]bilirubin). Rats were killed 10 or 30 min after the start of the bilirubin infusion. Each brain was homogenized, and subcellular fractions were isolated by high-speed gradient centrifugation in sucrose media. The gradients were separated into aliquots of 2 mL, and the protein content was determined in each aliquot. Radioactivity was determined by scintillation counting, and the content of bilirubin per milligram of protein was calculated. Statistical comparisons were performed with Kruskal-Wallis nonparametric ANOVA. There were highly significant differences in bilirubin content per milligram of protein among subcellular compartments in all groups and at both time points. In all groups there were relatively high concentrations of bilirubin in the myelin fraction, an interesting observation in light of the theory that membranes are the primary target of bilirubin toxicity. The very high concentration of bilirubin relative to protein in cytoplasm, ribosomes, and mitochondria in the hyperosmolar group are also notable in light of data from hyperbilirubinemic animals in which changes in electrophysiology or energy metabolism only appeared after hyperosmolar opening of the blood-brain barrier. The present data may be useful in planning in vitro studies of bilirubin toxicity in cell organelles.

Bilirubin oxidation in brain

Bilirubin is a product of heme catabolism which by virtue of its lipid solubility can cross the blood-brain barrier and enter the brain. Neonatal jaundice is a common transitional phenomenon which is due to the combination of increased heme catabolism and rate limitations as far as hepatic conjugation and biliary excretion of bilirubin. In the great majority of cases this is an innocuous condition, which is even posited to have some beneficial effects due to the ability of bilirubin to quench free oxygen radicals. However, because bilirubin is neurotoxic, hyperbilirubinemia in the newborn may exceptionally result in death in the neonatal period, or survival with severe neurological sequelae (kernicterus). Bilirubin enters the brain through an intact blood-brain barrier. Clearance of bilirubin from brain partly involves retro-transfer through the blood-brain barrier, and possibly also through the brain-CSF barrier into CSF. Work in our lab during the past 5 years has substantiated earlier work which had suggested that bilirubin may also be metabolized in brain. The responsible enzyme is found on the inner mitochondrial membrane, and oxidizes bilirubin at a rate of 100-300 pmol bilirubin/mg protein/minute. The enzyme activity is lower in the newborn compared with the mature animal, and is also lower in neurons compared with glia. Studies of different rat strains have documented genetic variability. The enzyme is cytochrome-c-dependent, but has as yet not been unequivocally identified. The rate of oxidation of bilirubin is such that this enzyme probably contributes meaningfully to the clearance of bilirubin from brain.

Oxidation of bilirubin in the brain-further characterization of a potentially protective mechanism

Bilirubin is a well-known neurotoxin and presents a particular problem in newborn infants. This is partly due to the high incidence of unconjugated hyperbilirubinemia in that age group, but may also be due to increased vulnerability to bilirubin toxicity. The brain may be able to protect itself against bilirubin toxicity through a process of oxidation. The responsible enzyme is localized on the inner mitochondrial membrane and appears to be more active in glia than in neurons and to increase in activity with postnatal maturation. Here we have investigated the possibility that the responsible enzyme might be a cytochrome oxidase, malate dehydrogenase, or monoamine oxidase, all enzymes located on the inner mitochondrial membrane. Mitochondria were obtained from rat brains through homogenization and differential centrifugation in sucrose medium. The ability of mitochondrial membranes to oxidize bilirubin was measured by following the change in optical density at 440 nm of a bilirubin solution to which a membrane suspension had been added. The activity was not changed by in vitro inhibitors of malate dehydrogenase or monoamine oxidase, but was moderately inhibited by ketoconazole and clotrimazole, both known inhibitors of hepatic cytochrome P450 oxidases. Activity was inhibited by depletion of cytochrome c in the mitochondria and reconstituted by reintroducing cytochrome c into the reaction mixture. The reaction was not modified by the addition of a free radical quencher, but was inhibited by removal of oxygen from the reaction mixture. The activity was significantly inhibited by cyanide. Activity was retained in a 100,000-g pellet and was not influenced by the addition of NAD, NADP, NADH, NADPH, GSH, or GSSH to this pellet. We conclude that the bilirubin-oxidizing activity in brain mitochondrial membranes is cytochrome c dependent, but does not appear to be unequivocally identifiable as a cytochrome P450 oxidase.

Brain bilirubin content is increased in P-glycoprotein-deficient transgenic null mutant mice

P-glycoprotein (P-gp), encoded by the mdr1a gene, is an ATP-dependent plasma membrane protein that is expressed in abundance on the blood-brain barrier (BBB). P-gp limits the CNS influx and retention of a variety of lipophilic compounds. We hypothesized that brain bilirubin content after an i.v. bilirubin infusion would be increased in P-gp-deficient mdr1a null mutant transgenic mice (mdr1a(-/-)) compared with controls. Eighteen mdr1a(-/-) null mutant and 18 P-gp-sufficient wild type mice (+/+) were anesthetized and 50 mg/kg bilirubin infused through the tail vein. Brain bilirubin content (mean +/- SEM) 10 min after infusion was significantly higher in mdr1a(-/-) (18.1 +/- 2.4 nmol/g) compared with (+/+) mice (10.4 +/- 1.0 nmol/g). Brain bilirubin content declined 60 min after infusion but remained higher in mdr1a(-/-) (10.3 +/- 1.4 nmol/g) compared with (+/+) mice (5.3 +/- 0.9 nmol/g). Brain bilirubin clearance did not differ between groups (t 1/2 approximately 55 min). We conclude that P-gp-deficient mdr1a(-/-) mice have significantly higher brain bilirubin content compared with controls after an i.v. bilirubin load. These data suggest that 1) bilirubin is a substrate for P-gp and 2) the increased brain bilirubin content in mdr1a(-/-) mice is due to enhanced brain bilirubin influx. We speculate that BBB P-gp provides a protective effect against bilirubin neurotoxicity by reducing brain bilirubin influx.

Oxidation of bilirubin by rat brain mitochondrial membranes-genetic variability

Bilirubin is oxidized by brain mitochondrial membranes at a rate which may contribute significantly to clearance of bilirubin from brain. Different strains of congenitally jaundiced rats (Gunn rats) vary widely as far as the mortality rate of the homozygous (jaundiced) pups. Because the ability to oxidize bilirubin in brain may protect against toxicity, we hypothesized that the ability to oxidize bilirubin would be lower in Gunn rat strains (ACI/N-j) with a high mortality rate in the homozygous pups. Mitochondria were obtained from young rat brains by differential centrifugation in sucrose gradients. The mitochondria were ruptured by sonication. The change in optical density of a bilirubin solution at 440 nm was measured over time following addition of the membrane suspension. The rate of bilirubin oxidation was significantly lower in rats of the RHA/N-j strain both at 7-8 days of age and in adults, compared to rats of the ACI/N-j and the Sprague-Dawley strains at the same age points. Differences in mortality rates between the RHA/N-j and the ACI/N-j strains of Gunn rats could not be explained on the basis of differences in the ability of brain mitochondrial membranes to oxidize bilirubin, as these activities were lower in the RHA/N-j rats, which also have lower mortality rates, but higher in the ACI/N-j rats, which have remarkably high mortality rates. This study also confirmed previous findings relative to age maturation of the enzyme activity.

Effect of bilirubin on erythrocyte shape and haemolysis, under hypotonic, aggregating or non-aggregating conditions, and correlation with cell age

The effect of unconjugated bilirubin on the morphology and haemolysis of human erythrocytes was accomplished under distinct incubation conditions: (i) hypotonic medium, with bilirubin concentrations ranging from 1 x 10(-9) to 1 x 10(-4) mol l-1; (ii) isotonic medium, with 171 mumol l-1 bilirubin, in the absence of albumin (aggregating conditions), using non-separated and age-fractionated erythrocytes; (iii) isotonic medium, with 171 mumol l-1 bilirubin, in the presence of a surplus of human serum albumin (non-aggregating conditions), and using sulfisoxazole as a bilirubin displacer (bilirubin/albumin and sulfisoxazole/ albumin molar ratios of 0.5 and 4.0, respectively). Our data showed that low concentrations of bilirubin (1 x 10(-7) to 1 x 10(-5) mol l-1) protect against hypotonic haemolysis and induce crenation, while higher bilirubin concentrations induce haemolysis and lead to membrane disruption. When aggregating conditions were used, these phenomena were reproduced, the younger cells being significantly more susceptible to crenation while the older erythrocytes showed increased susceptibility to haemolysis. In non-aggregating conditions, haemolysis was virtually absent, though crenation was evident. Based on the above observations we conclude that the first step of erythrocyte bilirubin toxicity is crenation due to an expansion of the outer membrane leaflet by bilirubin mono-anion location. This effect is more evident in younger cells and explains the protection against the hypotonic haemolysis. Insertion of bilirubin deeper into the bilayer, facilitated by higher concentrations (> or = 1 x 10(-4) mol l-1) and cell age, produces an unstable situation, where bilirubin acid aggregation is apparently the main cause for haemolysis and cell destruction.

Therapeutic approaches to neonatal jaundice: an international survey

Jaundice is one of the most common clinical phenomena in the neonatal period and a frequent indication for treatment with phototherapy, exchange transfusion, or drugs. The present study documents the variability in approaches to the treatment of this condition. A mail questionnaire was sent to neonatal units worldwide. One hundred and eight answers (49% response rate) were received from Europe (n = 72), North America (n = 28), Africa (n = 7), and Asia (n = 1). The neonatal intensive care units represented by the respondents had 31 +/- 18 beds [mean +/- SD], and 638 +/- 519 admissions per year. All units offered phototherapy, 106/108 performed exchange transfusion, while 44/108 used some form of drug therapy. There was considerable variability among the units in their approaches to the jaundiced neonate. This applied to all aspects of care, including type of phototherapy lights used, practical implementation of phototherapy, use of fluid supplementation, and use of prophylactic phototherapy. The majority used written protocols for investigation and treatment of neonatal jaundice and would let their decision on whether to treat be influenced by the infant's clinical state. There was great variability between units in the level of serum bilirubin that would trigger therapy. This applied across weight groups and to phototherapy as well as exchange transfusion. The significant heterogeneity in our approach to the treatment of jaundiced neonates suggests that our understanding of the biology of neonatal jaundice is inadequate and that further research will be necessary in order to provide a more solid biological foundation for therapy.