Abstract
Objective
To assess whether a lower maximum dose of soybean oil lipid emulsion reduces bilirubin neurotoxicity measured by brainstem auditory evoked response wave V latency.
Study design
Single-center Bayesian randomized trial (January 2021-January 2024) in a level IV neonatal intensive care unit. Infants born at <27 weeks or ≤750 g were stratified by phototherapy regimen, randomized to 1.5 vs 3.0 g/kg/day maximum soybean oil lipid emulsion (Intralipid) doses through day 14.
Results
Of 143 eligible infants, 134 were enrolled (67 lower dose, 65 higher dose). Mean [SD] daily lipid administration was 1.1 [0.3] vs 1.96 [0.58] g/kg/day. Unbound bilirubin levels were similar between groups although values ≥ 30 and ≥ 40 nM occurred more often in the higher dose group (14.5% vs 28.6% and 6.5% vs 16.1%), even at low total serum bilirubin levels (3–8 mg/dL). Adjusted wave V latency did not differ between groups (7.36 ms [0.53] vs 7.22 ms [0.57], aMD: −0.14 [95% CrI −0.34, 0.06]), with no interaction with phototherapy dose. Clinical outcomes and growth were comparable.
Conclusions
The lower maximum lipid dose did not reduce brainstem auditory evoked response wave V latency. Elevated unbound bilirubin levels occurred at the higher dose, suggesting the need for a multicenter safety trial in extremely preterm infants.
Trial registration
ClinicalTrials.gov https://clinicaltrials.gov/study/NCT04584983
Early parenteral nutrition with lipid emulsion is recommended1 for extremely preterm infants (born at <28 weeks’ gestation) for maintenance energy requirements. Soybean oil lipid emulsion (Intralipid)2 is commonly initiated after birth; however, the optimal dose for extremely preterm infants remains unclear.1,3 Stepwise increases in the soybean oil lipid emulsion to a maximum target of 3 g/kg/day are recommended2 to avoid adverse potential effects, including bilirubin neurotoxicity. Although 3 g/kg/day is the usual maximum dose in our center and other centers, this dose may increase unbound free fatty acids and displace sufficient bilirubin from albumin for unbound bilirubin to cross the blood barrier and cause bilirubin neurotoxicity,4–6 even at total serum bilirubin levels considered safe.4,5,7
We conducted a pragmatic, randomized safety trial to compare a lower maximum parenteral lipid dose (1.5 g/kg/day) with the recommended maximum dose (3 g/kg/day) in extremely preterm infants during the first 14 postnatal days, the period of highest vulnerability for bilirubin neurotoxicity. Our primary outcome was brainstem auditory evoked response (BAER) wave V latency (eFigure 1; available at www.jpeds.com)8 which is increased by reversable or lasting bilirubin neurotoxicity.9–11 Based on prior research in our center,12 we hypothesized that the higher dose would result in higher mean unbound bilirubin and free fatty acid levels but similar BAER wave V latency (primary outcome) with no interaction with phototherapy dose. Secondary outcomes included predischarge mortality, morbidity, and growth.
Methods
This parallel group superiority trial with factorial design was performed at a 118-bed level IV academic neonatal intensive care unit in Children’s Memorial Hermann Hospital, Houston, Texas. The institutional review board at McGovern Medical School approved the protocol,13 (eDocument 1) and the study team obtained informed parental consent for all participating infants. This study followed the Consolidated Standards of Reporting Trials reporting guidelines.
Eligible infants were <27 weeks’ gestation or ≤750 g birth-weight, <36 hours old, inborn, lipid-naïve, and enrolled in the National Institutes of Child Health and Human Development Neonatal Research Network Cycled Phototherapy Trial (NCT03927833).
Randomization
Web-based software (REDCap) was used to randomize infants 1:1 to a lower or higher maximum lipid dose in permuted block sizes of 4 to 6 after stratification by treatment group in the Cycled Phototherapy Trial, as part of a factorial design to assess whether cycled phototherapy modified the effects of lipid dose.14 The allocation sequence was known only to the study statistician (C.P.). The BAER expert and statistician were blinded to the treatment group. Other investigators, the clinical team, and parents were unblinded.
Intervention
The lower dose was started at 0.5 g/kg/day of a 20% soybean oil lipid emulsion and advanced by 0.5 g/kg/day each succeeding day to a maximum of 1.5 g/kg/day, a dose that prevents essential fatty acid deficiency.15 The higher dose was started at 1 g/kg/day and advanced by 1 g/kg/day to a maximum of 3 g/kg/day. The clinical team monitored serum triglycerides daily. In both groups lipid dosing was not advanced at values > 250 mg/dL and was reduced at values > 400 mg/dL. Lipids emulsions received photoprotective covering and delivered at a constant rate as low as 0.1 ml/h to deliver over 24 hours whenever possible for at least 7 days and reduced when enteral feeding volumes exceeded 80 ml/kg/day.
Primary Outcome Assessment
Investigators (L.H. or E.R.) assessed BAERs using standardized methods.16 The protocol included the following parameters: 80 dB, 100 ms clicks, with a rate of 20.1 clicks per second, filtered from 30–3000 Hz with ≥2000 repetitions averaged (three rarefaction and one condensation bilaterally), and electrode impedances <1 kilo-ohm. For unsatisfactory recordings at 80 db, 90 db was attempted, a level that if unable to ascertain is classified as a profound hearing deficit.
BAER assessment was routinely scheduled at 36 postmenstrual age (PMA) when the assessment may first be reliably assessed. For a minority of patients, BAERs were not successfully assessed until a later PMA because of unavoidable electrical interference in a neonatal intensive care unit environment. Because wave V latency is expected to decrease with increasing PMA, a study statistician (K.L.) used a predefined method to schedule BAERs after 36 weeks PMA to minimize any difference between groups in mean PMA at testing.17
Wave V latency was assessed by a blinded independent expert (S.S.) and adjusted for PMA at testing during the analyses. For infants with an absent wave V latency after serial testing, the criteria in eTable 1 (available at www.jpeds.com) were used to differentiate a technically unsatisfactory recording from one indicating severe bilirubin neurotoxicity that not only greatly increases the latency but also decreases the wave amplitude sufficiently for the wave to disappear.8 The few infants whose absent wave V latency was attributed to severe bilirubin neurotoxicity were conservatively assigned a wave V latency of 9.59 ms, a value 10% greater than maximum latency value recorded in any study infant with a discernible wave V latency.
Secondary Outcome Measures
Prespecified measures included other BAER measures that may be affected by bilirubin neurotoxicity,9,18 serum unbound bilirubin (mean, highest, and values ≥ 30 and ≥ 40 nM (that may increase risk of bilirubin neurotoxicity),19 free fatty acid levels (mean and highest), and albumin assessed using a fluorescent probe (Fluoresprobe Laboratories).5,20 Before assessing trial results, unbound bilirubin levels were changed from a primary to a secondary outcome because of uncertainty about the levels likely to be associated with bilirubin toxicity.21 Blood samples were obtained daily for six consecutive days starting the day after attaining the maximum goal lipid dose for each group. We also assessed the highest total serum bilirubin and direct bilirubin reported from our hospital laboratory (on blood samples collected within 4 hours of study blood sampling) as well as the clinical diagnoses and growth outcomes. Follow-up evaluation at 2 years adjusted age (post-term) is in progress.
Post hoc exploratory outcomes included an ordinal outcome comprised of death and the number of severe morbidities: Grade 3 bronchopulmonary dysplasia (BPD),22 grade III-IV intraventricular hemorrhage, periventricular leukomalacia, late onset sepsis, surgical necrotizing enterocolitis, stage 2 or greater retinopathy of prematurity, or presence of plus disease.
Statistical Analysis
Assuming no interaction between lipid and phototherapy dose, our sample size of ≥60 infants per group was selected to provide ≥80% power at an alpha of 0.05 to identify a decrease in wave V latency of >0.3 ms with the lower lipid dose assuming a SD of 0.56 ms, a 13% mortality rate, and an 8% rate of technically unsatisfactory BAERs. A 0.3 ms difference is approximately equal to the expected decrease in latency per week in gestational age between 25 and 36 weeks’ gestation.23,24 The data safety monitoring board recommended termination of the trial after interim analyses of the first 65 infants.
Bayesian analyses were performed to identify the probability of treatment benefit, promote nuanced conclusions, and avoid the limitations of frequentist analyses.25–27 For binary and ordinal outcomes, a neutral prior probability was centered at an OR of 1 and a 95% credible interval (CrI) of 0.25–4.0. For continuous outcomes, a neutral prior was centered on a mean difference (MD) of 0 between groups, with an SD equal to the observed SD of each outcome, corresponding to a prior 95% CrI of approximately ±2 SDs around 0. Logistic, linear, and ordinal logistic models were used in assessing the RR, OR, or MD and 95% CrIs adjusted for stratification variable. The likelihood that the lower dose resulted in any reduction in wave V latency at the PMA at testing was estimated. We also estimated the probability of the minimum clinically important difference28 between groups in their wave V latency: 0.18 ms. This value is half the increase in wave V latency between conservative and aggressive phototherapy in a center in which conservative phototherapy was associated with both higher total serum bilirubin values and adverse outcomes at 2 years among 501–500g birthweight infants.9
Because an absent wave V due to severe bilirubin neurotoxicity may be misinterpreted as technical difficulties, we performed post hoc sensitivity analyses in which infants designated as technically unsatisfactory were treated as indicative of severe bilirubin neurotoxicity.
To determine whether the effect of lipid treatment on wave V latency and rate of unbound bilirubin values ≥ 30 nM differed by phototherapy treatment assignment, we evaluated an interaction term in our primary analyses comparing the lipid and phototherapy groups. Evidence for interaction was assessed based on whether the posterior distribution of the interaction coefficient was centered near the null value and whether the 95% CrI included 0. All analyses were performed using the BRMS package in R software version 4.2.1 (R Foundation for Statistical Computing).
Results
Between January 8, 2021, and January 7, 2024, we enrolled 134 (93.7%) of 143 eligible infants; 67 were randomized to each dosage group (Figure 1). Consent was withdrawn for two higher dose infants before lipid administration. Baseline characteristics were similar between groups (Table I).
Figure 1.
CONSORT diagram.
Table I.
Baseline maternal and infant characteristics by parenteral lipid group
| Lipid group | ||
|---|---|---|
| Characteristic | Lower dose (n = 67) | Higher dose (n = 65) |
| Maternal age (y), mean (SD) | 28.6 (6.1) | 28.7 (6.2) |
| Maternal ethnicity, n (%) | ||
| Hispanic | 22 (33%) | 23 (35%) |
| Non-Hispanic | 37 (55%) | 40 (62%) |
| Unknown | 8 (12%) | 2 (3%) |
| Maternal race, n (%) | ||
| White | 37 (55%) | 35 (54%) |
| Black or African American | 20 (30%) | 25 (38%) |
| Asian | 1 (2%) | 1 (2%) |
| Not specified or unknown | 9 (13%) | 2 (3%) |
| More than one race | 0 (0%) | 2 (3%) |
| Singleton pregnancy, n (%) | 45 (67%) | 46 (71%) |
| Cesarean delivery, n (%) | 53 (79%) | 56 (86%) |
| Antenatal steroids, n (%) | 60 (90%) | 62 (95%) |
| Magnesium sulfate, n (%) | 55 (82%) | 52 (80%) |
| Gestational age by obstetric estimate (weeks), mean (SD) | 24.8 (1.6) | 25.1 (1.7) |
| Birthweight (g), mean (SD) | 737 (180) | 711 (179) |
| Male sex, n (%) | 33 (49%) | 36 (55%) |
| 1-minute Apgar score, mean (SD) | 5 (2) | 5 (2) |
| 5-minute Apgar score, mean (SD) | 7 (2) | 7 (2) |
| Birth head circumference (cm), mean (SD) | 22.1 (1.9) | 21.9 (2.0) |
| Birth length (cm), mean (SD) | 31.1 (3.1) | 31.0 (2.6) |
| Birthweight z score*, mean (SD) | 0.1 (0.9) | −0.1 (1.1) |
| Birth head circumference z score*, mean (SD) | −0.3 (0.9) | −0.6 (1.4) |
| Birth length z score*, mean (SD) | −0.4 (1.2) | −0.5 (1.3) |
| Age at randomization (h), mean (SD) | 23.1 (5.2) | 23.0 (4.6) |
| SNAPPE II score, mean (SD) | 66 (16) | 65 (17) |
| Phototherapy dose | ||
| Cycled, n (%) | 33 (49%) | 32 (49%) |
| Continuous, n (%) | 34 (51%) | 33 (51%) |
Mean (SD); n (%).
Fenton, 2013 PMID: 23758808. SNAPPE II, Score for Neonatal Acute Physiology with Perinatal Extension-II.
Lipid Administration
As detailed in eTable 2 (available at www.jpeds.com) the mean parenteral lipid exposure over days 1–7 and days 1–14 was 1.20 (0.24) g/kg/day and 1.10 (0.31) g/kg/day, respectively, in the lower dose group and 1.97 (0.64) g/kg/day and 1.96 (0.58) g/kg/day in the higher dose group. Because lipid administration was often interrupted or reduced for serum triglycerides >250 mg/dL, particularly in the higher dose group, the lipid dose administered was highly variable with a mean value each day less than the maximum goal (eFigure 2; available at www.jpeds.com).
Primary Outcome
We assessed BAERs in 115 survivors excluding 17 infants whose deaths prevented testing. (n = 10 lower dose, n = 7 higher dose) and two higher dose infants whose parents withdraw consent before any lipid administration. Our primary analyses to assess the effect of lipid dose on wave V latency included 110 survivors (56 lower dose; 54 higher dose) after excluding 5 lower dose infants: 1 with Kleeblattschadel syndrome and acquired CMV infection and 4 infants (all higher dose) designated as having technically unsatisfactory recordings and no discernible wave V latency. We included two infants (one lower dose and one higher dose) designated as having no identifiable wave V latency due to bilirubin neurotoxicity. The mean (SD) PMA at testing at testing was 38.4 (4.2) weeks with the lower dose vs 39.3 (5.6) weeks with the higher dose.
There was no evidence of a meaningful interaction between lipid dose and mode of phototherapy on wave V latency (coefficient: 0.01, 95% CrI: −0.37, 0.39) or rate of unbound bilirubin values ≥ 30 (coefficient: 0.32, 95% CrI: −0.63, 1.31). The adjusted wave V latency (n = 110) in the lower dose group (mean [SD], 7.36 ms [0.53]) was not less than in the higher dose group (7.22 ms [0.57]) (MD −0.14 [95% CrI −0.34, 0.06)] (Table II), providing no evidence of bilirubin neurotoxicity in the higher dose group. The probability of any true increase in wave V latency was low in both the primary and sensitivity analyses (8% and 56%, respectively). The probability of any difference in latency between groups exceeding the minimum clinically important difference was also low in both the primary and sensitivity analyses (35% and 17%).
Table II.
Primary outcomes of infants by parenteral lipid dose group
| Lipid group | Adjusted mean difference (95% CrI)* | Probability of a greater latency with the higher lipid dose† | Probability of a difference between groups in latency greater than the minimal clinical difference‡ | ||
|---|---|---|---|---|---|
| Lower dose (n = 67) | Higher dose (n = 65) | ||||
| Primary analyses (n = 110)§ | (n = 56)§ | (n = 54)§ | – | – | – |
| Wave V latency, ms, mean (SD) | 7.36 (0.53) | 7.22 (0.57) | MD: −0.14 (−0.34, 0.06) | 8% | 35% |
| Unable to measure wave V latency, n (%) | 1 (2%) | 4 (7%) | – | – | – |
| Sensitivity analyses (n = 114)¶ | (n = 56)¶ | (n = 58)¶ | – | – | – |
| Wave V latency, ms, mean (SD) | 7.36 (0.53) | 7.38 (0.82) | MD: 0.02 (−0.23, 0.28) | 56% | 17% |
Mean (SD); n (%).
The lower dose group minus the higher dose; adjusted for randomization stratification variables and postmenstrual age at testing.
The posterior probability that the latency of the higher dose group minus the latency of the lower dose group is greater than zero.
Values less −0.18 or greater than +0.18 were selected to be the minimum clinically important difference as values less than half the difference in latency between conservative and aggressive phototherapy in a center in which conservative phototherapy was associated with higher total serum bilirubin values and adverse outcomes at 2 years (PMID 22289854).
Seventeen infants died before BAER testing. Seven had no measurable wave V latency; five were excluded for technical limitations, and two were included with imputed maximum latency × 1.10 (9.59 ms). See eTable 1.
n = 114 includes 4 additional higher-dose patients included in sensitivity analyses of wave V latency (primary analysis, n = 110); values were imputed as maximum × 1.10 (9.59 ms).
Secondary and Exploratory Outcomes
Other BAER wave measurements that may be affected by bilirubin neurotoxicity were comparable between lower and higher lipid groups (eTable 3; available at www.jpeds.com). Unbound bilirubin values were highly variable with similar means in lower vs higher dose infants (13.20 nM [7.83] vs 12.29 nM [7.76], MD 0.77 [95% CrI −1.96, 3.48]) although worrisome values ≥ 40 nM were observed in 6.5% of lower vs 16.1% (relative risk 0.59 [95% CrI 0.26, 1.29]) of higher dose infants (Table III). The correlation between unbound bilirubin and total serum bilirubin was low (R = 0.4), and some infants had unbound bilirubin levels ≥40 nM at total serum bilirubin values of 3–8 mg/dL (eFigure 3; available at www.jpeds.com). However, wave V latencies adjusted for PMA in infants with unbound bilirubin ≥40 nM were similar to those with unbound bilirubin <40 nM (7.27 ms [1.33] vs 7.25 ms [0.45]).
Table III.
Secondary and post hoc exploratory outcomes by parenteral lipid dose group
| Lipid groups | Adjusted mean difference or relative risk (95% CrI)‡ | ||
|---|---|---|---|
| Laboratory values | Lower dose (n = 63)† | Higher dose (n = 57)† | |
| Mean unbound bilirubin, nM, mean (SD) | 13.20 (7.83) | 12.29 (7.76) | MD: 0.77 (−1.96, 3.48) |
| Highest measured unbound bilirubin ≥30 nM, n (%) | 9 (14.5%) | 16 (28.6%) | RR: 0.62 (0.34, 1.11) |
| Highest measured unbound bilirubin ≥40 nM, n (%) | 4 (6.5%) | 9 (16.1%) | RR: 0.59 (0.26, 1.29) |
| Highest measured unbound bilirubin, nM, mean (SD) | 22.02 (17.70) | 23.41 (17.53) | MD: −1.55 (−7.83, 4.69) |
| Age at highest measured unbound bilirubin, days, mean (SD) | 5.8 (2.5) | 6.7 (3.0) | MD: −0.78 (−1.77, 0.19) |
| Non-esterified fatty acids, mM§, mean (SD) | 0.39 (0.33) | 0.58 (0.44) | MD: −0.2 (−0.34, −0.06) |
| Unbound free fatty acids, nM§, mean (SD) | 30.98 (34.15) | 41.36 (27.64) | MD: −10.5 (−21.74, 0.71) |
| Highest measured unbound free fatty acids, nM§, mean (SD) | 63.77 (81.32) | 106.28 (94.02) | MD: −42.94 (−75.82, 9.05) |
| Albumin, g/dL, mean (SD)‖ | 2.7 (0.3) | 2.6 (0.3) | MD: −0.06 (−0.05, 0.16) |
| Hospital laboratory values | Lower dose (n = 63)† | Higher dose (n = 57)† | Adjusted Mean difference or relative risk (95% CrI)† |
| Highest measured total bilirubin, mg/dL, mean (SD) | 5.2 (1.3) | 5.0 (1.8) | MD: 0.1 (−0.4, 0.7) |
| Age at highest measured total bilirubin, days, mean (SD) | 5.7 (2.3) | 6.2 (2.8) | MD: −0.4 (−1.3, 0.5) |
| Highest measured direct bilirubin, mg/dL, mean (SD) | 1.0 (1.2) | 1.0 (1.1) | MD: −0.02 (−0.4, 0.4) |
| Direct bilirubin ≥1.5 mg/dL, n (%) | 10 (16%) | 9 (16%) | RR: 1.0 (0.5, 1.9) |
| Age at highest measured direct bilirubin, days, mean (SD) | 5.3 (2.6) | 6.8 (3.0) | MD: −1.5 (−2.5, −0.4) |
| Lipid groups | |||
| Hospitalization outcomes | Lower dose (n = 67) | Higher dose (n = 65) | Adjusted estimate (95% CrI)‡ |
| Death before discharge, n (%) | 10 (15%) | 7 (11%) | RR: 1.2 (0.6, 2.7)* |
| NICHD BPD grade¶# n (%) | – | – | OR: 0.8 (0.4, 1.4) |
| None | 12 (21%) | 9 (16%) | |
| Grade 1 | 30 (52%) | 27 (47%) | |
| Grade 2 | 10 (17%) | 18 (31%) | |
| Grade 3 | 6 (10%) | 4 (7.4%) | |
| Severe IVH# n (%) | 7 (12%) | 12 (20%) | RR: 0.7 (0.4, 1.4) |
| PVL# n (%) | 2 (3.5%) | 4 (6.9%) | RR: 0.8 (0.3, 2.1) |
| NEC stage II + n (%) | 16 (27%) | 25 (42%) | RR: 0.7 (0.5, 1.1) |
| Surgical NEC | 6 (10%) | 5 (8.5%) | RR: 1.1 (0.5, 2.6) |
| Late onset sepsis# n (%) | 16 (26%) | 22 (36%) | RR: 0.8 (0.5, 1.2) |
| Severe ROP# n (%) | 4 (7%) | 6 (10%) | RR: 0.8 (0.3, 1.9) |
| Death or severe morbidity (surgical NEC, BPD grade 3, severe ROP, LOS, PVL, severe IVH), n (%) | 32 (48%) | 38 (58%) | RR: 0.9 (0.6, 1.1) |
| Death or count of severe morbidities, n (%) | – | – | OR: 0.8 (0.5, 1.4) |
| Survival no morbidity | 35 (52%) | 27 (42%) | |
| Survival with 1 morbidity | 15 (22%) | 20 (31%) | |
| Survival with 2 morbidities | 3 (4.5%) | 5 (7.7%) | |
| Survival with 3 morbidities | 4 (6%) | 6 (9.2%) | |
| Death | 10 (15%) | 7 (11%) | |
| Death or any BPD, n (%) | 56 (84%) | 56 (86%) | RR: 1.0 (0.9, 1.1) |
| Death or severe IVH, n (%) | 15 (22%) | 18 (28%) | RR: 0.9 (0.5, 1.4) |
| Death or PVL, n (%) | 12 (18%) | 11 (17%) | RR: 1.0 (0.6, 2.0) |
| Death of surgical NEC, n (%) | 14 (21%) | 11 (17%) | RR: 1.2 (0.6, 2.2) |
| Death of late onset sepsis, n (%) | 22 (33%) | 26 (40%) | RR: 0.9 (0.6, 1.3) |
| Death or severe ROP, n (%) | 14 (21%) | 13 (20%) | RR: 1.0 (0.6, 1.9) |
| Length of stay, birth to discharge, days, mean (SD) | 128 (60) | 133 (69) | MD: −4.8 (−33.3, 24.7) |
| Growth outcomes | Lower dose (n = 67) | Higher dose (n = 65) | Adjusted mean difference (95% CrI)‡ |
| 14 postnatal days (end of intervention) weight, g¶ mean (SD) | 785 (177) | 743 (175) | 33 (−215, 278) |
| 36 weeks postmenstrual age¶** | – | – | – |
| Weight, g, mean (SD) | 2104 (288) | 1995 (383) | 69.9 (−560, 695) |
| Length, cm, mean (SD) | 41.2 (2.7) | 40.9 (3.2) | 0.5 (−0.6, 1.6) |
| Head circumference, cm, mean (SD) | 29.6 (1.5) | 29.1 (1.7) | 0.6 (−0.02, 1.2) |
| Discharge¶ | – | – | – |
| Weight, g, mean (SD) | 3407 (1070) | 3383 (1060) | 19 (−1173, 1241) |
| Length, cm, mean (SD) | 48.7 (5.0) | 48.5 (4.5) | 0.2 (−1.6, 1.9) |
| Head circumference, cm, mean (SD) | 34.4 (3.4) | 34.0 (2.3) | 0.4 (−0.6, 1.5) |
| Days to regain birthweight, days†† | 5 (7) | 5 (7) | 0.3 (−2.3, 2.8) |
| Weight gain during intervention period, birth to 14 postnatal days, g/day¶, mean (SD) | 1.7 (4.9) | 2.3 (6.1) | −0.6 (−2.6, 1.4) |
| Weight gain from birth to 28 postnatal days, g/day¶, mean (SD) | 8 (5) | 8 (4) | −0.03 (−1.6, 1.5) |
| Growth during birth to 36 weeks postmenstrual age¶** | – | – | – |
| Weight, g/day, mean (SD) | 18 (3) | 17 (4) | 0.4 (−0.9,1.7) |
| Length, cm/week, mean (SD) | 0.9 (0.3) | 0.9 (0.2) | 0 (−0.09, 0.08) |
| Head circumference, cm/week, mean (SD) | 0.7 (0.2) | 0.7 (0.1) | 0 (−0.05, 0.06) |
Mean (SD); n (%). BPD, bronchopulmonary dysplasia; FFAu, unbound free fatty acids; IVH, intraventricular hemorrhage; NEC, necrotizing enterocolitis; NICHD, National Institute of Child Health and Human Development; PVL, periventricular leukomalacia; ROP, retinopathy of prematurity.
A 27% Bayesian posterior probability of increased death at discharge with the lower compared to the higher lipid dose group.
Infants received no blood draws (total n = 12 infants) including for not reaching maximum lipid dose per group (n = 8) and early death (n = 4), and therefore, are excluded from these results.
The higher dose as the reference; mean difference is lower dose minus higher dose; adjusted for randomization stratification variables.
Infant (n = 1) missing value as blood sample not enough volume to measure unbound free fatty acids.
Missing data from deceased infants, excluding deceased infants (n = 17: n = 7 in the higher dose, n = 10 in the lower dose).
Missing data from clinical values, excluded from analysis. BPD missing n = 23 (of these, n = 16 died, n = 2 transferred to another hospital prior to 36 weeks, n = 5 discharged home prior to 36 weeks). Severe IVH grade III or higher missing n = 6 (n = 6 died and had no HUS). PVL before 28 postnatal days missing n = 4 (n = 4 died and had no HUS). Culture positive late onset sepsis n = 1 (n = 1 died). Severe ROP n = 13 (n = 13 died and had no exam documented).
Excluding infants who discharged before 36 weeks PMA (n = 6: n = 2 in the higher dose, n = 4 in the lower dose).
Missing data from 1 infant that did not regain birthweight by 28 postnatal days, excluded from analysis.
Infants received no clinically indicated hospital laboratory albumin measures that corresponded with research blood draw days (n = 15 in higher dose, n = 16 in lower dose).
Like unbound bilirubin, unbound free fatty acid values also were highly variable with comparable mean (SD) values for lower vs higher dose groups (30.98 nM [34.15] vs 41.36 nM [27.64]) although the highest values tended to be higher in the higher dose group (63.77 nM [81.32] vs 106.28 nM [94.02]).
No apparent differences between groups were observed for all other measured laboratory values, birth hospitalization outcomes, hospital stay, or growth.
Discussion
Our randomized trial investigated the concern posited for 50 years4,29 that a commonly used parenteral lipid dose (3 g/kg/day) may increase bilirubin neurotoxicity. Such a dose may increase unbound free fatty acids and displace sufficient bilirubin from albumin for unbound bilirubin to cross the blood barrier and cause bilirubin neurotoxicity,4,5 and alter the BAERs.9–11 Indeed, the displacement of bilirubin from albumin is established as potent with soybean oil lipid emulsion as with sulfisoxazole,6 a well-documented cause of bilirubin neurotoxicity.30
Our analyses of wave V latency provided no clear evidence that, compared with a lower dose, the higher lipid dose caused bilirubin neurotoxicity identifiable from BAER assessment when the lipid dosage was limited as in our study to maintain serum triglycerides <250 mg/dL. Because the lower and higher dose groups had similar average levels of unbound bilirubin and unbound free fatty acids, it is unsurprising that wave V latencies were comparable between groups.
However, unbound bilirubin levels were highly variable, and a larger study would be required to exclude the possibility that the higher lipid dose causes bilirubin neurotoxicity in a small subset of our study sample. Based on a recent analysis relating unbound bilirubin levels at five postnatal days to outcomes at 2 years in 1101 extremely low birth weight infants,19 8% of infants treated with aggressive phototherapy across centers of the Neonatal Research Network may be expected to have neurodevelopmental impairment associated with elevated unbound levels. Centers that administer higher average or maximum soybean oil lipid emulsion doses than in our trial would be expected to have higher unbound bilirubin values and a greater risk of bilirubin neurotoxicity than at our center. Thus, the higher lipid dose may increase unbound bilirubin levels and the risk of bilirubin neurotoxicity in some centers, depending on potential risk modifiers such as rates of hypoxic/ischemic events, intracranial hemorrhage, infections and administration of heparin or medications that may displace bilirubin from albumin.4 Soybean oil lipid emulsion and soy-medium chain triglyceride-olive fish oil lipid emulsion produce similar increases in free fatty acids and unbound bilirubin but not total serum bilirubin with escalating doses of lipid from 1 to 3 g/kg/day. It is particularly concerning that phototherapy did not significantly reduce unbound bilirubin in infants receiving lip emulsions at these doses. Because phototherapy isomerizes only albumin-bound bilirubin, it can prevent further increases in unbound bilirubin only indirectly and gradually by slowly reducing bilirubin production and therefore will not be recognized with monitoring of total serum bilirubin levels as in current clinical practice.
The strengths of our study include the importance of the question investigated, randomization of patients to treatment groups, measurement of unbound free fatty acids and unbound bilirubin using a state-of-the-art methods,5 blinded ascertainment of the primary outcome, the performance of research-grade BAER assessments in the largest sample of extremely preterm infants to date, the use of Bayesian analyses to assess the probability of a difference considered to be clinically important,25 and the ongoing evaluation of neurodevelopmental outcomes at 2 years adjusted age (to be reported separately when completed).
Our study’s limitations include a limited sample size, some uncertainty about how to interpret an absent wave V latency in small number of infants, the possibility that bilirubin toxicity may not affect the BAERs in all infants, and the absence of unbound bilirubin and unbound free fatty acid measurements when triglycerides were elevated. Most importantly, our single center design is likely to limit the generalizability of our findings and potentially provides unwarranted assurance of the safety of conventional lipid dosing in other centers.
In conclusion, compared with the commonly recommended higher parenteral lipid dose, the lower lipid dose did not reduce bilirubin neurotoxicity as assessed by BAER wave V latency in our single center study. However, worrisome unbound bilirubin values did occur, particularly at the higher dose, that may cause bilirubin neurotoxicity in a subgroup of patients, and this risk may be greater in other centers using similar dosing strategies. Our findings, in combination with other studies, suggest that a large multicenter trial is needed to verify the safety and value of commonly used doses of soybean oil lipid emulsion or other parenteral lipid emulsions administered to extremely preterm infants in different centers.
Supplementary Material
Declaration of Competing Interest
L.H. reports support for the present manuscript from the National Institutes of Health (NIH) Center for Clinical and Translational Sciences (NCATS) KL2 Career Development Award (5KL2TR003168) and the NIH Loan Repayment Program (L40HD109887). No other relationships or activities that could appear to have influenced the submitted work were reported.
We would like to acknowledge members of the data safety and monitoring board, including Dr. Sarah N. Taylor, MD, MSCR, and Dr. Jeffrey M. Maisels, MD. We would further like to acknowledge Michelle Christi, RN, NNP; Gabriella Dominguez, RN; Ileana Hernandez, BS; Apoorva Mahatme, MPH; Ilse Reyna, MS; Emily Stephens, RN; and Jaleesa Wade, RN, for their support in the conduct of the trial. We would also like to thank every parent of an infant enrolled in this study for partnering with us to determine the best way to provide care for some of our most vulnerable patients.
Glossary
- BAER
Brainstem auditory evoked response
- CrI
Credible interval
- MD
Mean difference
- PMA
Postmenstrual age
Footnotes
CRediT authorship contribution statement
Lindsay F. Holzapfel: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Jon E. Tyson: Writing – review & editing, Supervision, Resources, Methodology, Investigation, Conceptualization. Steven M. Shapiro: Writing – review & editing, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation. Claudia Pedroza: Writing – review & editing, Methodology, Formal analysis, Data curation. Eric Reynolds: Writing – review & editing, Investigation, Data curation. Kyung Hyun Lee: Writing – review & editing, Methodology, Formal analysis, Data curation. Matthew A. Rysavy: Writing – review & editing, Validation, Investigation, Formal analysis. Alan Kleinfeld: Writing – review & editing, Investigation, Data curation. Andrew H. Huber: Writing – review & editing, Methodology, Data curation. Mar Romero Lopez: Writing – review & editing, Investigation. Amir M. Khan: Writing – review & editing, Investigation. Cody Arnold: Writing – review & editing, Visualization, Validation, Supervision, Methodology, Investigation, Formal analysis, Conceptualization.
The current Trial registration is actually this: https://clinicaltrials.gov/study/NCT04584983.
References
- 1.Robinson DT, Calkins KL, Chen Y, Cober MP, Falciglia GH, Church DD, et al. Guidelines for parenteral nutrition in preterm infants: the American Society for Parenteral and Enteral Nutrition. J Parenter Enteral Nutr 2023: 830–858. 10.1002/jpen.2550 [DOI] [PubMed] [Google Scholar]
- 2.Fresenius Kabi USA LLC. Intralipid® (fat emulsion for intravenous use) prescribing information. Accessed March 19, 2026. https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=
- 3.Embleton ND, Jennifer Moltu S, Lapillonne A, van den Akker CH, Carnielli V, Fusch C, et al. Enteral nutrition in preterm infants (2022): a position Paper from the ESPGHAN Committee on nutrition and invited experts. J Pediatr Gastroenterol Nutr 2023;76:248–68. 10.1097/MPG.0000000000003642 [DOI] [PubMed] [Google Scholar]
- 4.Amin SB, Maisels MJ, Watchko JF. Early lipid infusions and unbound bilirubin in preterm neonates: a cause for concern? J Pediatr Mosby Inc 2017;184:6–7. 10.1016/j.jpeds.2017.01.065 [DOI] [PubMed] [Google Scholar]
- 5.Hegyi T, Kleinfeld A, Huber A, Weinberger B, Memon N, Shih WJ, et al. Effects of soybean lipid infusion on Unbound free Fatty acids and Unbound bilirubin in preterm infants. J Pediatr 2017;184:45–50.e1. 10.1016/j.jpeds.2016.12.026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hegyi T, Huber A, Oh W, Kleinfeld A. Unbound free fatty acids from intralipid displace bilirubin from albumin, comparable to sulfisoxazole. Pediatr Res 2025. 10.1038/s41390-025-04673-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Amin SB. Effect of free Fatty acids on bilirubin–albumin binding affinity and unbound bilirubin in premature infants. J Parenter Enteral Nutr 2010;34:414–20. 10.1177/0148607110362529 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Hall JW. In: Dragin S, ed. New handbook auditory evoked responses. 3rd ed. Boston, MA: Pearson Education; 2006. [Google Scholar]
- 9.Lasky RE, Church MW, Orlando MS, Morris BH, Parikh NA, Tyson JE, et al. The effects of aggressive vs. conservative phototherapy on the brainstem auditory evoked responses of extremely-low-birth-weight infants. Pediatr Res 2012;71:77–84. 10.1038/pr.2011.17 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Shapiro SM, Popelka GR. Auditory impairment in infants at risk for bilirubin-induced neurologic dysfunction. Semin Perinatol 2011;35.:62–170. 10.1053/j.semperi.2011.02.011 [DOI] [PubMed] [Google Scholar]
- 11.Wennberg RP, Ahlfors CE, Bickers R, McMurtry CA, Shetter JL. Abnormal auditory brainstem response in a newborn infant with hyperbilirubinemia: improvement with exchange transfusion. J Pediatr 1982;100:624–6. 10.1016/S0022-3476(82)80771-3 [DOI] [PubMed] [Google Scholar]
- 12.Arnold C, Tyson JE, Pedroza C, Carlo WA, Stevenson DK, Wong R, et al. Cycled phototherapy dose-finding Study for extremely low-birth-weight infants: a randomized clinical trial. JAMA Pediatr 2020;174:649–56. 10.1001/jamapediatrics.2020.0559 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Holzapfel LF, Arnold C, Tyson JE, Shapiro SM, Reynolds EW, Pedroza C, et al. Effect of reduced versus usual lipid emulsion dosing on bilirubin neurotoxicity and neurodevelopmental impairment in extremely preterm infants: study protocol for a randomized controlled trial. BMC Pediatr 2023;23:347. 10.1186/s12887-023-04149-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hegyi T, Chefitz D, Weller A, Huber A, Carayannopoulos M, Oh W, et al. Factors affecting the relationship between total and unbound bilirubin in preterm and term infants. Acta Paediatr Int J Paediatr 2025;114:1977–81. 10.1111/apa.70067 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gramlich L, Ireton-Jones C, Miles JM, Morrison M, Pontes-Arruda A. Essential fatty acid requirements and intravenous lipid emulsions. J Parenter Enteral Nutr 2019;43:697–707. 10.1002/jpen.1537 [DOI] [PubMed] [Google Scholar]
- 16.Shapiro SM, Riordan SM. Review of bilirubin neurotoxicity II: preventing and treating acute bilirubin encephalopathy and kernicterus spectrum disorders. Pediatr Res 2020;87:332–7. 10.1038/s41390-019-0603-5 [DOI] [PubMed] [Google Scholar]
- 17.Pocock SJ, Simon R. Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial. Biometrics 1975;31: 103–15. [PubMed] [Google Scholar]
- 18.Gelineau-Morel R, Usman F, Shehu S, Yeh HW, Suwaid MA, Abdulsalam M, et al. Predictive and diagnostic measures for kernicterus spectrum disorder: a prospective cohort study. Pediatr Res 2024;95:285–92. 10.1038/s41390-023-02810-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Arnold CC, Maric I, Wong RJ, Tyson JE, Stevenson DK. Unbound bilirubin and risk of severe neurodevelopmental impairment in extremely low birthweight newborns. Pediatr Res 2025;98:166–74. 10.1038/s41390-025-03872-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Hegyi T, Kleinfeld A, Huber A, Weinberger B, Memon N, Shih W, et al. Unbound bilirubin measurements by a novel probe in preterm infants. J Maternal-Fetal Neonatal Med 2019;32:2721–6. 10.1080/14767058.2018.1448380 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Jensen EA, Dysart K, Gantz MG, McDonald S, Bamat NA, Keszler M, et al. The diagnosis of bronchopulmonary dysplasia in very preterm infants an evidence-based approach. Am J Respir Crit Care Med 2019;200: 751–9. 10.1164/rccm.201812-2348OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kaga K, Hashira S, Marsh RR. Auditory brainstem responses and behavioural responses in pre-term infants. Br J Audiol 1986;20:121–7. 10.3109/03005368609079005 [DOI] [PubMed] [Google Scholar]
- 23.Scaioli V, Brinciotti M, Capua MDI, Lori S, Janes A, Pastorino G, et al. A multicentre database for normative brainstem auditory evoked potentials (BAEPs) in children: methodology for data collection and evaluation. Open Neurol J 2009;3:72–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Amrhein V, Greenland S, McShane B. Scientists retire statistical significance. Nature 2019;567:305–7. 10.1038/d41586-019-00857-9 [DOI] [PubMed] [Google Scholar]
- 25.Goligher EC, Heath A, Harhay MO. Bayesian statistics for clinical research. Lancet 2024;404:1067–76. 10.1016/S0140-6736(24)01295-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lee JJ, Harrell FE, LaVange LM, Spiegelhalter DJ. Embracing bayesian methods in clinical trials. JAMA 2026;335.:1664–6. 10.1001/jama.2026.4179 [DOI] [PubMed] [Google Scholar]
- 27.McGlothlin AE, Lewis RJ. Minimal clinically important difference. JAMA 2014;312:1342. 10.1001/jama.2014.13128 [DOI] [PubMed] [Google Scholar]
- 28.Bryan H, Shennan A, Griffin E, Angel A. Intralipid-its rational use in parenteral nutrition of the newborn. Pediatrics 1976;58:787–90. [PubMed] [Google Scholar]
- 29.Andersen DH, Blanc WA, Crozier DN, Silverman WA. A difference in mortality rate and incidence of kernicterus among premature infants allotted to two prophylactic antibacterial regimens. Pediatrics 1956;18:614–25. [PubMed] [Google Scholar]
- 30.Wijeysundera DN, Austin PC, Hux JE, Beattie WS, Laupacis A. Bayesian statistical inference enhances the interpretation of contemporary randomized controlled trials. J Clin Epidemiol Elsevier USA 2009;62(1):13–21.e5. 10.1016/j.jclinepi.2008.07.006 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.