Effect of Free Fatty Acids on Bilirubin–Albumin Binding Affinity and Unbound Bilirubin in Premature Infants

Abstract

Background

The author has previously shown that intravenous lipid intake may be associated with an increase in unbound bilirubin in infants ≤28 weeks gestational age. The objective of this study was to evaluate whether this increase in unbound bilirubin is mediated by free fatty acids and to examine the secondary effect of free fatty acids on bilirubin–albumin binding affinity.

Methods

A prospective study was conducted to include 26 infants ≤32 weeks gestational age with indirect hyperbilirubinemia and receiving intravenous lipids during the first 10 postnatal days. Blood samples were collected for unbound bilirubin, binding affinity, and free fatty acid measurement at varying intravenous lipid intakes (1–3 g/kg/d). Regression analyses were performed to evaluate the roles of free fatty acids and binding affinity as mediators.

Results

Intravenous lipid intake was significantly associated with an increase in free fatty acids and unbound bilirubin in infants ≤28 weeks but not >28 weeks gestational age. In infants ≤28 weeks gestational age, each unit increase in free fatty acids was significantly associated with a decrease in binding affinity, which was significantly associated with an increase in unbound bilirubin.

Conclusions

In infants ≤28 weeks gestational age, intravenous lipid intake may be associated with an increase in unbound bilirubin, and this is mediated by an increase in free fatty acids and a secondary decrease in binding affinity. In infants >28 weeks gestational age, higher intravenous lipid intake may be used because it is unassociated with increases in free fatty acids and unbound bilirubin.

There is increasing scientific evidence that unbound or free bilirubin (B_f) may be a better predictor of bilirubin-induced neurotoxicity than total serum bilirubin (TSB).^1,2 The B_f concentration depends not only on TSB and albumin concentration but also on the binding affinity of bilirubin to albumin. Several factors may influence bilirubin–albumin binding affinity, including elevated free fatty acid (FFA) concentrations.^1,3-5 Intravenous lipid emulsion (IL), commonly used to provide a concentrated isotonic source of calories during the first few weeks of postnatal life, may be associated with elevated FFA concentration in premature infants because premature infants clear FFA less effectively than term infants.^6-8

The concern for a potential bilirubin displacing effect that may be associated with increased IL intake secondary to elevated FFA concentration often leads to limited IL intake in premature infants with indirect hyperbilirubinemia, regardless of gestational age (GA).^9,10 The limited evidence that supports the restriction of IL intake during indirect hyperbilirubinemia in premature infants is based mainly on indirect estimation of B_f using the reserve bilirubin-binding capacity as measured by fluorometer or saturation index or direct estimation of B_f by the peroxidase method but using a single peroxidase concentration.^11-14 The use of a single peroxidase concentration may result in underestimation of B_f secondary to sample dilution, and therefore, the peroxidase method was recently modified to use 2 or more concentrations of peroxidase.¹⁵ Using the modified peroxidase method, we recently demonstrated that higher IL intake may be associated with increased B_f concentration in premature infants ≤28 weeks GA but not in those >28 weeks GA.¹⁶ Despite evidence from clinical studies that establishes an association between IL intake and B_f concentration, the underlying mechanism that may help to establish a causal relationship has not been well elucidated using appropriate statistical methods.^11,14,16 Moreover, the reasons for the influence of GA on the association between IL intake and B_f concentration have not been investigated. Therefore, our primary objective was to evaluate whether the association between B_f concentration and IL intake is mediated by FFA concentration and the secondary effect of FFA on bilirubin–albumin binding affinity in premature infants. Our secondary objective was to evaluate the influence of GA on the role of FFA concentration as a mediator for the effect of IL intake on B_f concentration in premature infants.

Methods

A prospective nested study was performed to determine whether the effect of IL intake on B_f is mediated by FFA in premature infants with indirect hyperbilirubinemia using the validated modified peroxidase method.¹⁷ The study was approved by the institutional research review board. Informed consent was obtained from the parents of the study patients.

Study Population

Premature infants ≤32 weeks GA who were delivered at the University of Rochester Medical Center between June 2006 and June 2008 and enrolled in a National Institutes of Health-funded study entitled “Bilirubin-Induced Auditory Neuropathy in Preterms” and receiving parenteral nutrition including IL were eligible for this study. GA was assessed by obstetrical dating criteria or, when obstetrical data were inadequate, by Ballard examination. Infants with major craniofacial malformations, chromosomal disorders, TORCH (toxoplasmosis, other infections, rubella, cytomegalovirus, and herpes simplex virus) infections, family history of hearing loss, or those who were too clinically unstable for auditory testing between 24 and 48 hours after birth were excluded as part of the primary bilirubin study. In addition, infants with blood culture–positive sepsis and those receiving postnatal steroids during the first 10 postnatal days were excluded from this nested study. Because a previous study¹⁶ found that the association between IL and B_f was seen in premature infants ≤28 weeks GA but not in infants >28 weeks GA, patients were divided into 2 subgroups—(1) infants ≤28 weeks GA and (2) infants >28 weeks GA—to evaluate the influence of GA on the role of FFA as a mediator for the effect of IL intake on B_f concentrations.

The policy of the neonatal intensive care unit at Golisano Children's Hospital is to administer IL (20% Intralipid; Fresenius Kabi, Uppsala, Sweden) over 16 to 20 hours at a rate <0.15 g/kg/h.¹⁸ IL intake is gradually increased from 0.5 to 3 g/kg/d during the first week, at the discretion of the attending neonatologist.¹⁰ The major component fatty acids in Intralipid are linoleic (44%–62%), oleic (19%–30%), palmitic (7%–14%), linolenic (4%–11%), and stearic (1.4%–5.5%). IL intake is often limited to ≤2 g/kg/d in the presence of hyperbilirubinemia. The actual TSB concentrations at which IL was restricted were highly variable and dependent on the bias of the attending physician. IL is usually infused via central venous catheters in premature infants <30 weeks GA. The heparin that can affect lipid clearance is usually administered with central venous catheters at a dose of 0.5 unit/mL.^19,20 Phototherapy was used according to the institutional guidelines for the management of hyperbilirubinemia in premature infants that are currently based on TSB concentrations, birth weight, and the presence of clinical factors that increase the risk for bilirubin encephalopathy. Acidosis was evaluated as clinically indicated at the discretion of the attending neonatologist.

Outcome Variables

Blood samples were drawn on each enrolled patient on different days, while receiving varying IL intake between the third and tenth days after birth, to measure TSB (μmol/L) and B_f (nmol/L) by the peroxidase test (UB Analyzer; Arrows Company, Osaka, Japan). B_f was measured using UB Analyzer, a semiautomated spectrophotometer, approved by the US Food and Drug Administration.^21,22 The peroxidase test was performed at 2 peroxidase concentrations with the stock peroxidase diluted 1:42 or 1:83, and the steady-state B_f at each peroxidase concentration was used to calculate the equilibrium of free bilirubin (B_feq).¹⁵

The same aliquot of blood used to measure B_f was used to measure FFA concentration. Samples for FFA analyses were frozen and stored at –80°C until analysis. FFA concentrations were measured using an enzymatic analysis kit (Wako Chemicals USA, Richmond, VA) by the laboratory technician, blinded to IL intake and B_f concentration. The mean coefficient of variation for FFA concentration was 1%.

Serum albumin concentrations were measured twice between the second and fourth postnatal days on each patient by the bromocresol green method.²³ Because the half-life of serum albumin is about 3 weeks, albumin concentrations were assumed to be constant for the first 10 days. The highest B_feq on each IL intake, corresponding TSB, and the average of the 2 serum albumin concentrations were used to calculate the bilirubin–albumin equilibrium association binding constant (K, L/μmol) on each IL intake using the following equation:

$$\mathrm{K}=\frac{\mathrm{TSB}-{\mathrm{B}}_{\mathrm{f}}}{{\mathrm{B}}_{\mathrm{f}}(\text{Albumin}-\mathrm{TSB}+{\mathrm{B}}_{\mathrm{f}})}$$

The binding constant measures the binding affinity or strength of the bond between bilirubin and albumin.

Statistical Analyses

Statistical analyses were performed with STATA (version 10; Stata Corporation, College Station, TX). Analysis of variance with Scheffé multiple comparison tests was used to analyze continuous variables among multiple groups. If the Bartlett test for equal variances assumption was significant (P < .05), the Kruskal-Wallis test (nonparametric test) was used. A P value <.05 was considered significant. A series of regression models were built to evaluate whether FFA and binding affinity are mediators of the effect of IL on B_f.²⁴

Regression Models

FFA concentration and binding affinity were considered a priori mediators for the effect of IL on B_f as shown below:

$$\mathrm{IL}\text{intake}\to \text{increase in}\mathrm{FFA}\to \text{decrease in binding affinity}\to \text{increase in}{\mathrm{B}}_{\mathrm{f}}$$

Model 1. The principal predictor of interest is IL intake, the mediator is FFA, and the outcome is B_f, as shown below:

$$\mathrm{IL}\text{intake}\to \text{increase in}\mathrm{FFA}\to \text{increase in}{\mathrm{B}}_{\mathrm{f}}$$

For FFA to be a mediator, IL intake should be able to predict FFA molar concentration. The mediator (FFA) should be able to predict the outcome (B_f) in a regression model controlling for the principal predictor of interest (IL). Finally, addition of the mediator (FFA) to a multi-predictor model for the outcome (B_f) should attenuate the estimated coefficient for the predictor of interest (IL).

Model 2. The principal predictor of interest is FFA, the mediator is binding affinity, and the outcome is B_f, as shown below:

$$\text{Increase in}\mathrm{FFA}\to \text{decrease in binding affinity}\to \text{increase in}{\mathrm{B}}_{\mathrm{f}}$$

For binding affinity to be the mediator for the FFA effect on B_f, FFA should predict binding affinity and binding affinity should be able to predict the outcome in a model controlling for FFA (the predictor of interest), and addition of the mediator to a multipredictor model for the outcome should attenuate the estimated coefficient for FFA. To evaluate the influence of GA on the role of FFA and binding affinity as mediators, regression analyses were performed as a function of GA group.

Results

A total of 171 infants were eligible for the primary Bilirubin-Induced Auditory Neuropathy study between July 2006 and January 2008. Eight infants (6 with cranio-facial anomalies and 2 with TORCH infections) were excluded. Nine infants died or were critically ill within the first 36 hours and were not enrolled. A total of 118 infants were enrolled in the primary bilirubin study. Twelve infants (9 receiving hydrocortisone and 3 with active sepsis) were excluded from this nested study. Of the remaining 106 infants, 24 infants were >32 weeks GA, and 19 infants were started on enteral feedings within 36 hours after birth. Of the remaining 63 infants, 26 infants, aged 24 to 32 weeks GA and receiving parenteral nutrition and IL, were enrolled in this study. A total of 87 blood samples were drawn at varying IL intakes (average 3 samples from each patient) between the third and tenth day after birth and analyzed for B_f and FFA.

The mean GA at birth of the 26 infants was 27.7 weeks (median, 27.7 weeks). The mean birth weight of these infants was 1,064 g (median, 1,050 g). Sixteen of 26 (62%) infants were male, and 16 (62%) infants were Caucasian. One infant was small for GA. The mean albumin concentration was 2.8 g/dL (median, 2.8 g /dL; range, 2.2–3.7 g/dL). There was no significant difference in TSB and TSB/albumin molar ratio as a function of IL intake (Table 1). There was significant difference in FFA, FFA/albumin molar ratio, and B_f concentration as a function of IL intake (Table 1). None of the infants developed hyperglycemia and received insulin therapy during the study period. Although there was a difference in glucose infusion rate as a function of IL intake, there was no statistically significant association between glucose infusion rate and B_f concentration (P = .2). The strong association between IL intake and B_f remained statistically significant even after we controlled for GA and glucose infusion rate (P = .003). Among the 41 samples with prior evaluation for acidosis within 12 hours as part of the standard of care, only 5 samples were associated with prior acidosis (pH, < 7.25). The mean FFA concentration (1,797 μmol/L; n = 5) with the history of proximate acidosis was not statistically different from the mean FFA concentration (2,132 μmol/L; n = 36) associated with pH ≥ 7.25 (P = .68).

Table 1.

Clinical Characteristics of Study Patients (N = 26)

	Intravenous Lipid Emulsion Intake
	0.5 g/kg/d, (n = 18)	1 g/kg/d, (n = 22)	1.5 g/kg/d, (n = 16)	2 g/kg/d, (n = 12)	2.5 g/kg/d, (n = 7)	3 g/kg/d, (n = 12)	P
Gestational age, wk	27.7 ± 1.6	27.7 ± 1.8	27.6 ± 1.6	28 ± 2.5	27.4 ± 1.7	27.7 ± 1.6	.9
Glucose infusion rate, mg/kg/min	5.4 ± 2.4	6.6 ± 1.3	8.2 ± 2.1	8.5 ± 3	8.5 ± 2.5	8 ± 2.6	.002
TSB, mg/dL (μmol/L)	8.4 ± 1.8 (144 ± 31)	9.2 ± 2 (157 ± 34)	8.8 ± 2.8 (150 ± 48)	8.9 ± 1.8 (152 ± 31)	8.3 ± 2.1 (142 ± 36)	8.0 ± 2.0 (137 ± 34)	.6
TSB/albumin molar ratio	0.34 ± 0.07	0.38 ± 0.07	0.35 ± 0.1	0.35 ± 0.09	0.33 ± 0.08	0.33 ± 0.07	.5
FFA, μmol/L	1,210 ± 1,459	1,713 ± 1,764	1,383 ± 664	2,194 ± 1,536	2,992 ± 1,663	2,760 ± 1,262	.01
FFA/albumin molar ratio	2.9 ± 3.8	4.1 ± 4.2	3.3 ± 1.6	5.2 ± 3.9	7.2 ± 4.5	6.6 ± 2.9	.02
Bf, μg/dL	0.49 ± 1.8	0.82 ± 0.76	0.93 ± 1.0	1.01 ± 1.21	1.98 ± 2.39	2.11 ± 2.73	.02

Abbreviations: B_f, free or unbound bilirubin; FFA, free fatty acid; TSB, total serum bilirubin.

The results of regression analyses demonstrating the role of FFA as a mediator for the effect of IL intake on B_f concentration are shown in Table 2. IL intake predicts FFA molar concentration, and a mean increase in IL intake of 1 g/kg/d results in a mean increase in FFA by 649 mM. The mediator FFA also predicts the outcome B_f after controlling for the effect of IL on B_f. A mean increase in FFA of 1 mM results in an average increase in B_f by 0.0004 mM. Evaluation of a multi-predictor model for B_f with addition of the mediator FFA attenuates the effect of the IL on B_f with the decrease in the estimated coefficient from 0.62 to 0.33. These results indicate that FFA is, indeed, a mediator of the effect of IL on B_f. The results were similar when FFA/albumin molar ratio was used as a mediator instead of FFA (Table 2).

Table 2.

Regression Analyses to Evaluate FFA and FFA/Albumin Molar Ratio as Mediators for the Effect of IL Intake on B_f

Exposure Variable	Outcome Variable	Controlling Variable	Coefficient	P
IL	Bf		0.62	.001
FFA as a mediator
IL	FFA		649	.001
FFA	Bf	IL	0.0004	.000
IL	Bf	FFA	0.33	.04
FFA/albumin molar ratio as a mediator
IL	FFA/albumin molar ratio		1.5	.01
FFA/albumin molar ratio	Bf	IL	0.18	.000
IL	Bf	FFA/albumin molar ratio	0.34	.04

Abbreviations: B_f, free or unbound bilirubin; FFA, free fatty acid; IL, intravenous lipid emulsion; K, bilirubin–albumin binding affinity.

Table 3 demonstrates the role of bilirubin–albumin binding affinity as a mediator for the effect of FFA on B_f. FFA predicts binding affinity, with each unit increase in FFA leading to an average decrease in binding affinity of 0.009 L/mol. Binding affinity predicts B_f, when controlling for FFA. A decrease in binding affinity by 1 L/mol leads to an average increase in B_f by 0.022 mM. Upon addition of the mediator, which was binding affinity, to a multipredictor model for B_f, the effect of FFA on B_f was attenuated with the decrease in the estimated coefficient from 0.0005 to 0.0002. These results indicate that binding affinity is a mediator of the effect of FFA on B_f. The results were similar when FFA/albumin molar ratio was used as the primary predictor of interest instead of FFA concentration (Table 3).

Table 3.

Regression Analyses to Evaluate K as a Mediator for the Effect of FFA or FFA/Albumin Molar Ratio on B_f

Exposure Variable	Outcome Variable	Controlling Variable	Coefficient	P
FFA as the primary exposure of interest
FFA	Bf		0.0005	.000
FFA	K		–0.009	.000
K	Bf	FFA	–0.022	.000
FFA	Bf	K	0.0002	.001
FFA/albumin molar as the primary exposure of interest
FFA/albumin molar ratio	Bf		0.21	.000
FFA/albumin molar ratio	K		–4.1	.000
K	Bf	FFA/albumin molar ratio	–0.022	.000
FFA/albumin molar ratio	Bf	K	0.12	.001

Abbreviations: B_f, free or unbound bilirubin; FFA, free fatty acid; K, bilirubin–albumin binding affinity.

The influence of GA on the role of FFA and binding affinity as mediators for the effect of IL on B_f is shown in Table 4. For infants ≤28 weeks GA, there was a statistically significant association between IL intake and FFA or FFA/albumin molar ratio. FFA/albumin molar ratio also predicted bilirubin–albumin binding affinity. Binding affinity predicted B_f after controlling for FFA. For infants >28 weeks GA, IL intake was not significantly associated with FFA concentration or FFA/albumin molar ratio, and therefore, FFA or FFA/albumin molar ratio is not a mediator. There was a significant negative correlation between FFA and binding affinity. Similarly, there was significant negative correlation between binding affinity and B_f.

Table 4.

Influence of GA on the Role of FFA and K as Mediators for the Effect of IL Intake on B_f

GA Group	Exposure Variable	Outcome Variable	Coefficient	P
≤28 wk	IL	FFA/albumin molar ratio	2.6	.000
	FFA/albumin molar ratio	K	–4.1	.000
	K	Bf	–0.03	.000
>28 wk	IL	FFA/albumin molar ratio	0.83	.17
	FFA/albumin molar ratio	K	–3.4	.02
	K	Bf	–0.02	.000

Abbreviations: B_f, free or unbound bilirubin; FFA, free fatty acids; GA, gestational age; IL, K, bilirubin–albumin binding affinity.

Discussion

Indirect hyperbilirubinemia, which may lead to bilirubin-induced neurotoxicity, is a common condition in premature infants during the first postnatal week. Literature suggests that B_f is more closely associated with bilirubin-induced neurotoxicity than TSB.^25-29 Increased IL intake is essential for adequate nutrition and optimum growth of premature infants. However, because of the potential of bilirubin displacing the effect of FFA generated from IL, IL intake is often limited in premature infants with indirect hyperbilirubinemia, irrespective of GA.^6,9,10 We recently demonstrated that the amount of IL intake that may be tolerated without secondary increase in B_f is GA dependent.¹⁶ However, the underlying mechanism, specifically the role of FFA as a mediator for the effect of IL on B_f, was not elucidated. The findings of the current study provide causal explanation and strongly suggest that the effect of IL on B_f is mediated by elevated FFA concentration, which in turn leads to decreases in bilirubin–albumin binding affinity. In addition, our findings provide substantial evidence that the mediating effect of FFA is GA dependent and explain why increased IL intake may not be associated with an increase in B_f concentration in premature infants >28 weeks GA.

Bilirubin bound to albumin can be displaced by various anions, including FFA releasing B_f at a lower level of TSB concentration.¹ The relationship between FFA and the binding of bilirubin to albumin has been evaluated by several investigators.^3-5,14,30-32 FFA binds to albumin at 3 sites. The third binding site of FFA is also the highest affinity binding site of bilirubin to albumin. In vitro studies using crystalline human serum albumin have suggested that the first 2 mol of FFA preferentially bind to its first 2 binding sites. When the FFA/albumin molar ratio is ≥4:1, FFA competes with bilirubin at the primary site of bilirubin (also the third binding site of FFA), resulting in displacement of bilirubin. However, studies using neonatal plasma suggest that compared with crystalline human serum albumin, displacement of bilirubin from its primary site may occur at a lower FFA/albumin molar ratio.³³

Our results suggest that there is linear correlation between FFA/albumin molar ratio and B_f concentration in premature infants. Our findings are consistent with those of Spear et al,¹¹ who reported a linear correlation between B_f concentration measured using the hematofluorometer and FFA/albumin molar ratio, with infants who had an FFA/albumin molar ratio >4:1 having the largest increase in B_f. Spear et al¹¹ also reported that IL intake of 2 to 3 g/kg/d when infused slowly in premature infants <30 weeks GA may be associated with FFA/albumin molar ratio >4:1 and possible displacement of bilirubin from binding sites leading to elevated B_f levels. However, compared with the previous study by Spear et al,¹¹ we evaluated binding affinity to confirm the association between FFA and B_f. We found that as FFA/albumin molar ratio or FFA concentration increased, bilirubin–albumin binding affinity decreased, and a decrease in binding affinity correlated significantly with an increase in B_f. Also, according to our findings, the association between IL and FFA or FFA/albumin molar ratio appears to be GA dependent, with a significant association observed in infants ≤28 weeks GA but not those >28 weeks GA. Our findings are similar to the findings of Spear et al,¹¹ which demonstrated a significant increase in FFA/albumin molar ratio with each increase in lipid dosage for infants <30 weeks GA but not for infants ≥30 weeks GA in a study conducted before the routine use of antenatal steroids. In addition, increases in the FFA or FFA/albumin molar ratio had less of an effect on the bilirubin–albumin binding affinity in the group of infants >28 weeks GA compared with the group ≤28 weeks GA.

Our study differs from previous studies in the composition of the patient population. Previous studies included premature infants of higher GA (>27 weeks),^6,11-13,33 whereas we studied premature infants of low GA, the neonatal population with the greatest need for IL intake and also at the highest risk for bilirubin-induced neurotoxicity. The negative conclusions reached for the effect of IL on B_f by some previous studies may have been influenced by grouping infants >28 weeks GA with infants ≤28 weeks GA, possibly masking the significant association between IL intake and B_f concentration in more premature infants.^12,13

Our study also differs from previous work in that we statistically analyzed the role of mediators for the effect of IL on B_f as a function of GA and in the method of B_f measurement used. Previous work has involved either indirect estimation of B_f using a fluorometer or saturation index test or direct estimation of B_f using a peroxidase test with a single concentration of peroxidase.^6,11-13,33 A peroxidase test involving a single concentration of peroxidase may underestimate B_f.¹⁵ To minimize underestimation of B_f, we used a modified peroxidase test using 2 concentrations of peroxidase.^15,17 Recent reports suggest that measuring B_f at decreased sample dilution using the peroxidase test may further improve the accuracy of B_f measurement by decreasing the effect of dilution on intrinsic albumin binding of many ligands, including bilirubin.^15,34

One limitation of this study is that the findings are based on observational study. Clinical factors such as sepsis and steroids can explain the findings of elevated FFA; however, patients with active sepsis and postnatal steroid exposure were excluded.^35,36 The use of potential bilirubin displacers other than FFA may explain the findings; however, potential bilirubin displacers, such as sulfonamides, ceftriaxone, and ibuprofen, were not used during the study period.¹ Changes in binding affinity may also be due to allosteric/conformational changes in albumin or other bilirubin-binding proteins as a result of the clinical status of the infant.³⁷ Although our study has the limitations of an observational study, the statistical models used to analyze the role of FFA and binding affinity as mediators for the effect of IL intake on B_f provide novel yet powerful evidence for a biologically plausible mechanism and, therefore, a causal relationship between IL intake and B_f.²⁴ These findings are based on specific lipid preparation used during the study period and may not be applicable for lipid emulsions containing different FFA profiles.^12,38

In summary, our findings strongly suggest that increased IL intake may be associated with a significant increase in B_f concentration in premature infants ≤28 weeks GA and that this effect of IL on B_f is mediated by an increase in FFA concentration and a secondary decrease in binding affinity. Our findings also suggest that IL intake is not significantly associated with elevated FFA in infants >28 weeks GA, and therefore, higher IL intake is less likely to be associated with secondary increase in B_f in this patient population.