Unbound Bilirubin Measurements by a Novel Probe in Preterm Infants

Abstract

Background:

Hyperbilirubinemia occurs in over 80% of newborns and severe bilirubin toxicity can lead to neurological dysfunction and death, especially in preterm infants. Currently, the risk of bilirubin toxicity is assessed by measuring the levels of total serum bilirubin (TSB), which are used to direct treatments including immunoglobulin administration, phototherapy, and exchange transfusion. However, free, unbound bilirubin levels (Bf) predict the risk of bilirubin neurotoxicity more accurately than TSB.

Objective:

To examine Bf levels in preterm infants and determine the frequency with which they exceed reported neurotoxic thresholds.

Methods:

One hundred thirty preterm infants (BW 500–2000g; GA 23– 34 weeks) were enrolled and Bf levels measured during the first week of life by the fluorescent Bf sensor BL22P1B11-Rh. TSB and plasma albumin were measured by standard techniques. Bilirubin-albumin dissociation constants (Kd) were calculated based on Bf and plasma albumin.

Results:

Five hundred eighty samples were measured during the first week of life, with an overall mean Bf of 13.6± 9.0 nM. A substantial number of measurements exceeded potential toxic thresholds levels as reported in the literature. The correlation between Bf and TSB was statistically significant (r² 0.17), but this weak relationship was lost at high Bf levels. Infants < 28 weeks gestations had more hearing screening failures than infants ≥ 28 weeks gestation.

Conclusions:

Unbound (free) bilirubin values are very variable during the first week of life in preterm infants. A significant proportion of these values exceeded reported neurotoxic thresholds.

Introduction

Hyperbilirubinemia manifests in more than 80% of infants (1) and severe bilirubin toxicity can lead to neurodevelopmental handicap and death, especially in the preterm population. (2, 3). Currently, the risk of bilirubin toxicity is assessed by measuring the level of total serum bilirubin (TSB), which is used to direct treatment to reduce the risk of neurotoxicity.(4) The inability of TSB to identify the risk of handicap in a near-term or term infant is highlighted by the observation that over 1000 infants need to be treated with phototherapy in order to prevent one infant from reaching TSB levels that require an exchange transfusion. (5) The poor positive predictive value of TSB is matched by its suboptimal negative predictive value in preterm infants, who can demonstrate bilirubin neurotoxicity at low TSB levels.(6, 7)

Most bilirubin is bound by albumin in the serum and is therefore confined to the serum compartment. However, the minute unbound fraction of serum bilirubin (Bf) can cross the bloodbrain barrier and is responsible for neurotoxicity (8–10). The levels at which Bf exerts toxicity have not been clearly defined. Alhlfors used the peroxidase test to determine a Bf exchange transfusion threshold of 1.3 μg/dL /dL (22 nmol/L) per kilogram birth weight with a maximum allowable Bf of 4ug/dL (66 nmol/L).(10) However, infants with unbound bilirubin levels near or above 1 μg/ dL (17nM) may have subtle bilirubin-induced ABR wave latency and amplitude changes detected by ABR testing. (9) In a recent investigation by Amin, late preterm and term infants demonstrated auditory toxicity at Bf levels as low as 0.76 μg/dL (11 nM).(11)

Recently, we developed a Bf-specific probe through a process of iterative mutations and high throughput screening of fluorescently labeled fatty acid binding proteins and demonstrated that this probe can detect physiologic, nanomolar concentrations of Bf in plasma.(12) The goal of this investigation was to examine Bf levels, their correlation with TSB levels, and the frequency with which they exceed reported toxicity thresholds in preterm infants during the immediate postnatal period. (10, 11)

Methods:

Patients:

One hundred thirty infants with birth weight between 500 and 2000g, gestational age of 23 to 34 weeks were enrolled in this observational study. Data from 97 infants in this cohort have been reported previously to demonstrate the effect of increasing doses of soy-based lipid (Intralipid, IL) infusion on FFAu and Bf (13). Recruitment was carried out in the Neonatal Intensive Care Unit (NICU) at Rutgers Robert Wood Johnson Medical School (RWJMS) after parental informed consent was obtained. The study was approved by the IRBs of Rutgers and The Torrey Pines Institute for Molecular Studies.

Study Design:

Serum Bf and TSB levels were measured in preterm infants when clinicians decided that a laboratory test was indicated for the management of neonatal jaundice. Of note, Bf levels were measured in all available samples and most of TSB levels were accompanied by Bf, as permitted by sample availability. The number and timing of Bf and TSB samples were determined by the clinical staff. The total number of Bf levels were utilized for the comprehensive analysis of a Bf subgroup that was accompanied by TSB measurements used to examine the Bf/ TSB relationship.

Blood Sampling:

Infant blood samples were obtained during the first week of life and research samples were collected from residual blood drawn for clinical indications. These samples were collected immediately after completion of the requested clinical test, processed and frozen. De-identified samples were stored at −70⁰C and then shipped to Fluoresprobe Sciences for determination of Bf.

Bf and TSB measurements:

Bf was measured using fluorescently labeled mutants of fatty acid binding proteins.(12) The Bf probe (BL22P1B11-Rh) binds unconjugated bilirubin with high affinity (Kd=16 nM) but is poorly sensitive to FFA (Kd>3000 nM), conjugated bilirubin (Kd>300 nM), bilirubin photoisomers, bilirubin oxidation products, ibuprofen, and indomethacin. (12) Fluorescence was assessed at emission wavelengths of 525 and 580 nm for the Bf probe (excitation = 375 nm). The ratio of fluorescence at the two wavelengths, together with probe characteristics, yielded the Bf concentrations. Fluorescence was measured at 22°C using a handheld ratio fluorometer in which plasma sample volumes were 8 μL for Bf measurements and after dilution, the total volume was 200 μL so that Bf was measured at 25 fold dilution of the plasma sample.

The accuracy of the Bf probe measurements was confirmed previously by comparison with the peroxidase assay in adult plasma supplemented with bilirubin as well as in bilirubin-albumin in vitro measurements. (12) The Bf probe has an average CV of 3% for repeated measurements of aqueous bilirubin, bilirubin-albumin complexes and bilirubin spiked adult plasma over a Bf range from 1 to > 350 nmol/L. Measurements with the Bf probe in contrast to the peroxidase method determines the equilibrium Bf concentration directly in a single measurement and is insensitive to substances that can interfere with the peroxidase measurement.

TSB was measured using the diazo method at the RWJMS clinical laboratory. Plasma albumin was measured in infants prior to IL infusion using the Bromocresol Green (BCG) Albumin Assay (Sigma) method which was standardized with fatty acid-free human serum albumin (HSA) from 72 to 722 μM (CVs< 4%). Infant plasma (5 μL) was diluted into 200 μL of the BCG reagent and absorbance was measured at 620 nm. We calculated the Bilirubin –albumin equilibrium dissociation constants (Kd) in 53 infants, which is the inverse of the binding constant Ka using equation ${\text{K}}_{\text{d}}=\frac{{\text{B}}_{\text{f}}{\text{A}}_{\text{t}}}{\left(\text{TSB}-{\text{B}}_{\text{f}}\right)}-{\text{B}}_{\text{f}},$ where A^t is the measured albumin concentration and Kd is in moles/ L or nM.

Outcome data:

Brainstem audio-evoked responses were assessed using an Algo5 Newborn Hearing Screener (Natus Medical Inc., Pleasanton CA). The hearing assessment was part of the routine discharge protocol for all infants and not specifically for this study and was done between 35 and 37 weeks of gestational age. Failure was defined as “refer” on one or both ears at initial evaluation prior to NICU discharge.

Statistical analysis:

Changes in TSB and Bf as a function of demographic and clinical variables were analyzed by repeated measures analysis and verified by the generalized estimating equation semi-parametric regression method. Multiple regression analysis was performed to evaluate the correlation between TSB with Bf, taking into account confounding variables including birth weight, gestational age, gender, race, treatment interventions and 5 minute Apgar and the analysis was verified by non-parametric Spearman correlation. The analyses were performed with Statistica (StatSoft); p<0.05 was considered significant. Potential toxic thresholds were designated at Bf levels of 11, 17 and 22 nM, based on prior reports. (9–11) Finally, since in a previous report(13) we noted differential Bf responses among infants above and below 28 weeks of gestation, we divided the cohort into these two gestational age groups for additional analysis.

Results

The 130 study infants had a mean birth weight of 1345.2±430.7 g and gestational age of 29.4±3.2 weeks (Table 1). Five hundred-eighty Bf samples were obtained in this population at a mean age of 53.1±38.5 hours, with a significant positive correlation noted between Bf and age at sampling (r=0.273, p<0.05, Figure 1). The mean Bf level was 13.6±9.0 nM, with the 25^th percentile 8.5 nM and the 75^th percentile 16.3 nM. There were no differences in the initial Bf levels between male and female (6.6±2.5 nM vs. 6.4±2.1 nM) or black and white infants (6.2±2.1 nM vs. 6.5±2.0 nM). Of the total number of samples 173 (29.8%), 64 (11.0%) and 34 (5.9%) exceeded the reported neurotoxic thresholds of 11, 17 and 22 nM, respectively (14). The number of infants who had at least one level exceeding those prior thresholds was 54, 32 and 17, respectively.

Table 1:

Demographic Characteristics of Study Infants

Number of infants enrolled	130
Gestational Age (weeks)	29.4±3.2
Birthweight (g)	1345.2±430.3
Male, n (%)	66 (50.8)
Race/Ethnicity, n (%)	21 (15.3)
Asian	36 (26.3)
Black	71 (51.8)
White	2 (1.5)
Hispanic

Figure 1:

Age at Sampling and Unbound Bilirubin (Bf) Levels

Bf significantly correlated with TSB [r=0.411 (p<0.05)], but Bf levels in excess of 21 nM did not correlate with respective TSB levels (r=−0.113, p=n.s.). Maximum levels of Bf and TSB were noted on day of life 4, and the ratio of Bf to TSB was greatest on day five (Table 2).

Table 2.

Daily Unbound and Total Bilirubin Levels during the First Week of Life

Day	n	Bf (nM)	TSB (mg/dL)	Bf/TSB (nM × 10−4)
1	125	7.1±3.0	3.9±1.3	0.11±0.05
2	149	14.0±7.5	6.4±1.9	0.13+/0.06
3	130	15.2±7.6	6.2±2.5	0.16±0.09
4	95	17.6±12.5	6.3±2.1	0.19±0.15
5	46	16.7±12.6	5.2±1.9	0.21±0.15
6	11	13.5±5.0	5.5±2.2	0.16±0.07
≥7	23	15.4±6.9	5.1±2.7	0.21±0.09

Hearing screening data was available in 116 infants, with 20 failures. Comparing these two subgroups (pass and fail), we found significant differences in gestational age (29.9±3.1 vs. 28.3±3.0 weeks) and in birth weight (1413.9±418.0 vs. 1195.3±463.8 g). However, neither mean Bf nor TSB levels were significantly different between the two groups, and there were no differences in areas under the curve for Bf and TSB with respect to postnatal age. Maximum Bf levels were greater (25.0±24.5 vs. 20.8±12.5 nM) and maximum TSB levels lower (7.3±2.9 vs. 7.6±2.2 mg/dL) in infants who failed the screen, but these differences did not reach statistical significance (Table 3).

Table 3.

Hearing Screening Results

	All		< 28 weeks		# 28 weeks
Hearing screen results	Pass	Fail	Pass	Fail	Pass	Fail
N	96	20	27	9	69	11
TSB (mg/ dl) max	7.6±2.2	7.3±2.9	5.7±1.9	6.2±0.7	8.4±1.8	8.1±1.0
Bf (nM) max	20.8±12.5	25.0±24.5	27.2±19.8	19.0±5.9	18.3±6.8	21.9±21.1

Infants ≤ 28 weeks gestational age may be more susceptible to bilirubin-induced auditory toxicity since 25.0% of infants < 28 weeks and 13.8% of infants ≥ 28 weeks failed the hearing screening. Interestingly, in < 28 infants, peak Bf at levels of > 22, 17, and 11 nM identified 22.2%, 44.4% and 100%, and in infants > 28 weeks 27.2%, 45.5%, and 81.8% of infants with failed hearing screens.

Multiple linear regression was performed at each Bf sample with the following selected variables: weight at sample, gestational age, gender, race and 5 minute Apgar score. None of these variables had significant effects on the changes in Bf as a function of increasing TSB. We found that bilirubin-albumin binding is highly variable among infants independent of these demographic characteristics, with Kd values ranging from 15–75 nM (Figure 2).

Figure 2:

Kd Measurements

Discussion

Preterm infants are at high risk for poor outcomes due to a number of factors that include bilirubin toxicity. (15) As a consequence, this population generally receives aggressive phototherapy treatment (2) guided by serum TSB levels, based on the assumption that TSB concentrations possess a modicum of accuracy in predicting risk status from bilirubin toxicity. Unfortunately, the use of TSB leads to exposure of excessive numbers to phototherapy or exchange transfusion in order to protect a single term or near term infant from bilirubin toxicity. (5, 16) Conversely, TSB may underestimate the risk of toxicity, especially in the preterm infant, as exemplified by a report that peak TSB levels in infants with kernicterus were often less than half of the AAP threshold for exchange transfusions.(17) In an earlier investigation, autopsy examinations documented the presence of kernicterus in premature infants despite normal TSB levels. (18)

In contrast to TSB, there is substantial evidence that unbound, free bilirubin concentration (Bf) is a superior predictor of the risk of bilirubin toxicity, (19–21) which is especially important in the high-risk preterm infant population, who are often exposed to interventions that potentially compromise bilirubin binding to albumin. (22, 23) In the past, a number of methods were developed to measure Bf, but the only one to quantify it directly was the UB Analyzer (Arrows) that was based on the peroxidase assay.(24) This method had a number of limitations including the requirement of sample dilution, which resulted in severe underestimates of Bf in the presence of a competitor of bilirubin-albumin binding.(25, 26) In contrast, although we diluted 25 fold in our present study, a modified form of our newly developed Bf sensor will require no dilution and will be able to measure Bf accurately in whole blood in samples as small as 5μL compared to larger volumes for the UB Analyzer. (22) With preterm infants losing on average up to 30% of their blood volume weekly due to phlebotomy, this feature will be an essential one for clinical use in the NICU.

Our present studies provide insight into the role of Bf measurements in the management of highrisk preterm infants. In a similar investigation, van der Schoor and colleagues observed a large variation in Bf levels, as measured by the horseradish peroxidase reaction, and noted that it was independent of postnatal age, gender and risk factors.(6) They reported that Bf/TSB ratios were highest in small preterm infants, who demonstrated the lowest TSB levels. Similarly, we found that maximal Bf and TSB levels occurred on the fourth day of life. However, individual Bf levels were not predictable based on TSB alone, as confirmed by the wide range of albuminbilirubin binding constants we observed in these infants. This current report relates literature toxicity levels and peak Bf concentrations, whereas data on exposure to exogenous agents that affect bilirubin-albumin binding, such as Intralipid, are not included. These have been previously reported for a subset of this cohort and the lack of correlation between TSB and Bf at higher levels of Bf is very likely due to the high dose of Intralipid infusion. (23)

We found that premature infants frequently exceeded Bf levels that have previously been reported to be neurotoxic, with over 29 percent of our samples above 11 nM. (11) There was a significant correlation between Bf and TSB, but not at high Bf levels where this relationship would be the most clinically useful. Bilirubin neurotoxicity is thought to be common in premature infants below 28 weeks gestational age. (27) Consistent with this, infants in our < 28week group failed screening at almost twice the rate of those ≥ 28 weeks, with odds ratios for failure of 2.1 at Bf levels at Bf levels ≥ 11 nM, 1.8 at ≥ 17 nM, and 1.1 at ≥ 22 nM. However, we recognize the limitations of a single hearing evaluation and we plan to rectify this issue by analyzing brainstem auditory response waveforms and latencies in a future investigation. In summary, peak Bf levels in preterm infants may be associated with neurotoxicity, as evidenced by increased incidence of hearing failure. We acknowledge that our hearing data is limited by that fact that it is a retrospective collection and that no prospective power calculation was done. The results may be attributable to insufficient sample size, and no long-term follow-up of this cohort for hearing loss is available. However, it is clear that TSB levels are poorly predictive of outcomes, and correlate with Bf only at low levels. Moreover, bilirubin-albumin constants, which determine the ratios between Bf and TSB, are highly variable among infants. Our findings suggest that direct measurement of Bf will be proven to be a superior approach to determining when to start or continue therapy for hyperbilirubinemia in preterm infants.

Abbreviations:

TSB

total serum bilirubin

Bf

free bilirubin

BW

birth weight

GA

gestational age

Kd

bilirubin-albumin dissociation constant

IRB

institutional review board

CV

coefficient of variation

RWJMS

Robert Wood Johnson Medical School No reprints requested