Double-Blind, Randomized Trial of Long-Chain Polyunsaturated Fatty Acid Supplementation in Formula Fed to Preterm Infants
Objective. We tested the hypothesis that balanced addition of long-chain polyunsaturated fatty acid (LCPUFA) to preterm formula during the first weeks of life would confer long-term neurodevelopmental advantage in a double-blind, randomized, controlled trial of preterm formula with and without preformed LCPUFA.
Methods. The participants were 195 formula-fed preterm infants (birth weight <1750 g, gestation <37 weeks) from 2 UK neonatal units and 88 breast milk-fed infants. Main outcome measures were Bayley Mental Developmental Index (MDI) and Psychomotor Developmental Index (PDI) at 18 months and Knobloch, Passamanick and Sherrard’s Developmental Screening Inventory at 9 months’ corrected age. Safety outcome measures were anthropometry at 9 and 18 months, tolerance, infection, necrotizing enterocolitis, and death.
Results. There were no significant differences in developmental scores between randomized groups, although infants who were fed LCPUFA-supplemented formula showed a nonsignificant 2.6-point (0.25 standard deviation) advantage in MDI and PDI at 18 months, with a greater (nonsignificant) advantage (MDI: 4.5 points; PDI: 5.8 points) in infants below 30 weeks’ gestation. LCPUFA-supplemented infants were shorter than control infants at 18 months (difference in length standard deviation score: 0.44; 95% confidence interval: 0.08–0.8). No other significant short- or long-term differences in safety outcomes were observed. Breastfed infants had significantly higher developmental scores at 9 and 18 months than both formula groups and were significantly heavier and longer at 18 months than LCPUFA-supplemented but not control infants.
Conclusions. With the dose, duration, and preparation of LCPUFA used, efficacy was not demonstrated, although an advantage in later neurodevelopment cannot be excluded by global tests of development up to 18 months, particularly in infants below 30 weeks’ gestation. The surprising effect of LCPUFA-supplemented formula on growth 18 months beyond the intervention period needs to be confirmed in other studies using similar supplementation strategies. Additional follow-up of this cohort is critical at an age when more specific tests of cognitive function are possible.
Two groups of long-chain polyunsaturated fatty acids (LCPUFA) have received increasing interest recently: homologues of linoleic acid of the n-6 series (dihomogammalinolenic acid, arachidonic acid [AA]) and of α-linolenic acid of the n-3 series (eicosapentanoic acid, docosahexanoic acid [DHA]). These LCPUFA are synthesized from the precursor essential fatty acids principally by a process of chain elongation and desaturation. They are found in high concentrations in the phospholipids of cell membranes, notably in the central nervous system.1,2 In addition, AA, dihomogammalinolenic acid, and eicosapentanoic acid are precursors for eicosanoids,3 important modulators and mediators of a variety of essential biological processes.
Rapid accumulation of LCPUFA in the brain, particularly DHA, occurs from the third trimester to 18 months’ postpartum. Breast milk contains both the precursor essential fatty acids and adequate LCPUFA for structural lipid accretion. However, infant formulas traditionally have contained only the parent essential fatty acids, leaving the infant to synthesize LCPUFA endogenously. Both term and preterm infants who are fed formulas that contain minimal LCPUFA have been shown to have lower red cell LCPUFA4–6 and lower LCPUFA in the phospholipids of the cerebral cortex7,8 than infants who are fed breast milk. Whether these biochemical differences between formula-fed and breast milk-fed infants have clinical relevance in terms of long-term growth or neurodevelopment remains controversial and is the subject of a number of current studies. Although there is evidence that supplementing infant formulas with LCPUFA results in apparently favorable biochemical changes4,5,9 and, in some studies, transient or short-term neurocognitive effects,10,11 few data relate to long-term clinical outcome. However, primarily on the basis of biochemical data suggesting that preterm infants may have a reduced ability to synthesize LCPUFA from the parent essential fatty acids, European Society for Paediatric Gastroenterology and Nutrition12 recommended that formulas for low birth weight infants be enriched with metabolites of both linolenic and linoleic acids (DHA and AA) approximating to the levels typical of breast milk.
Since these recommendations, the stringency required for efficacy and safety testing of formula innovations has increased. A recent review by Simmer10 for the Cochrane Library concluded that no long-term benefit has been demonstrated for preterm infants who receive LCPUFA-supplemented formula. Moreover, there has been concern about possible adverse effects on growth.13 Thus, any study attempting to examine the case for adding LCPUFA to preterm infant formulas not only must prove an advantage for later outcome but also must be robust enough to provide reassurance on safety.
We conducted a large, prospective, randomized trial in an unselected population of preterm infants to test the hypothesis that the balanced addition of n-3 and n-6 LCPUFA to a currently available preterm infant formula during the initial period of hospitalization would result in improved subsequent neurodevelopmental outcome at 9 and 18 months’ corrected age. The parallel aim of the study was to test for safety and tolerance of the formula.
Preterm infants with birth weight <1750 g were recruited from 3 UK neonatal intensive care units in Nottingham (University Hospital and City Hospital) and Leicester. All had a gestational age <37 weeks and were free from congenital malformations known to affect neurodevelopment. Infants were eligible for the study if their mother had decided not to provide breast milk by 10 days of age and if they were tolerating enteral feeds at that time. The birth weight cutoff was chosen to increase the likelihood of the infant’s remaining on the trial formula for a minimum of 3 weeks. Infants received the trial formula until they were discharged from the neonatal unit. During their stay on the neonatal unit, infants were monitored intensively by research nurses who collected information each day on the infant’s clinical course, feed volumes (enteral, including the volume of each type of milk, and parenteral), and feed tolerance (see below). Weight was recorded daily and length and head circumference (HC) twice weekly when the infant’s clinical condition permitted. Obstetric data were collected at enrollment, together with information on social class (based on the UK Registrar General’s classification) and maternal educational achievements (recorded on a 5-point scale ranging from no qualifications to higher professional qualifications as described previously14).
Preterm infants whose mothers provided breast milk were recruited as a reference group and followed the same protocol for investigation and follow-up as formula-fed infants. Informed written consent was obtained, and the study protocol was approved by the ethics committees of the participating hospitals and the MRC Dunn Nutrition Unit, Cambridge.
The primary efficacy outcome was neurodevelopment at 18 months postterm, assessed using the mental and motor scales of the Bayley Scales of Infant Development II,15 from which are derived the Mental Developmental Index (MDI) and Psychomotor Development Index (PDI). These were calculated on the basis of postterm age. Secondary outcome measures were 1) neurodevelopment at 9 months, assessed using Knobloch, Passamanick and Sherrard’s Developmental Screening Inventory,16 which comprises 5 subscales (adaptive, gross motor, fine motor, language, and personal-social) and provide scores in weeks postterm age (a quotient was calculated for each subscale as [test score/postterm age in weeks] × 100; an overall quotient was calculated as the mean of the 5 subscale quotients); 2) neurologic impairment at 9- and 18-month follow-up (diagnosed by the examining pediatrician).
Key safety outcomes were growth (weight, length, and HC); infections (categorized as microbiologically confirmed skin sepsis, systemic infection with clinical and hematologic evidence only [high or low white cell count and/or low platelet count] and bacteriologically proven systemic infection [clinical and hematologic evidence with positive blood cultures]); necrotizing enterocolitis (NEC; categorized as suspected [clinical picture with abdominal distension] or confirmed [at surgery, postmortem, or by the presence of portal or intramural gas on abdominal radiograph]); hemorrhagic events (intracranial and pulmonary), requirement for respiratory support (days requiring >30% oxygen, days of mechanical ventilation) as an overall measure of illness; and major clinical events.
Tolerance outcomes were the number of days taken to reach full enteral feeds, number of stools passed per day, stool consistency (coded as soft, formed, or hard by nursing staff), and the presence of abdominal distension or diaper rash. At 9- and 18-month follow-up, the number of episodes of gastroenteritis (diarrhea with vomiting), upper respiratory tract infection, and chest infections requiring antibiotic treatment were recorded, together with the history or presence of eczema, wheeze, and asthma and information on the usual frequency and consistency of the infant’s stools.
Infants were randomized to receive either a preterm infant formula without additional LCPUFA (Prematil, Milupa: control formula) or a supplemented formula (Prematil with Milupan: LCPUFA-supplemented formula) containing a fat blend with vegetable oils (palm coconut, soya, sunflower) and milk fat with derivatives of linoleic and α-linolenic acid sourced from evening primrose oil (γ-linolenic acid) and egg lipids (AA and DHA). Formulas were provided in ready-to-feed form; both had identical appearance and smell. The composition of both formulas is shown in Table 1.
The allocation schedule for each center was generated using permuted blocks of randomized length by personnel who were not involved in subsequent aspects of the study. Infants were randomly assigned to receive 1 trial formula using dietary allocations stored in sealed opaque envelopes. Randomization was stratified by birth weight (<1200 g or >1200 g) to increase the likelihood that the smallest, sickest infants would be equally distributed between feed groups. Twins and triplets were randomized separately.
The 2 formulas were provided in color-coded containers. The code was held by the formula manufacturers and was not broken until the study and analyses were completed. Parents and study personnel were therefore blind to the dietary allocation throughout the study, follow-up, and data analysis periods.
The calculated sample size (100 infants per randomized group) permitted detection of a 0.4 standard deviation (SD) difference between diet groups with 80% power at 5% significance. Seventy-five subjects per randomized group (the approximate number seen at 18-month follow-up) would permit detection of a 0.46 SD difference between groups with 80% power at 5% significance.
Data for randomized groups were compared using Student’s t test for parametric and Mann-Whitney test for nonparametric data. Categorical variables were compared using χ2 or Fisher’s exact test. Multiple regression analysis was used to compare outcome between randomized groups after adjusting for potentially confounding variables. All principal analyses were performed on an intention-to-treat basis, without adjustment. Exploratory analyses were performed to examine outcome in relation to gestational age and the volume of trial formula consumed by the infant. Data from the breastfed reference group were compared with that from formula-fed infants using analysis of variance. Weight, length, and HC SD scores were calculated using British reference data.17
The trial profile is shown in Fig 1. A total of 195 infants were randomized, 100 to control formula (53 from center 1 and 47 from center 2) and 95 to LCPUFA-supplemented formula (50 from center 1 and 46 from center 2). Mean age at randomization was 5 (SD: 4) days for both groups. Trial diets were fed for a mean of 33 (SD: 17) days in control infants and 31 (SD: 21) days in LCPUFA-supplemented infants (P = .6). Seventy-four (84%) of the breastfed infants were recruited from center 1, and 14 (16%) were recruited from center 2.
Six control infants and 14 study infants were withdrawn from the study before completing 3 weeks on the trial diet, for the reasons shown in Fig 1; in 2 control infants (1 with NEC, 1 not tolerating feeds) and 2 LCPUFA-supplemented infants (both with NEC), the diet was considered a “possible” factor in the withdrawal, and in 2 additional LCPUFA-supplemented infants, the diet was thought to have “probably” contributed (these were twins whose mother was concerned about their poor weight gain and requested a change of diet). No other withdrawals were thought to be related to the infant’s diet. Between 3 weeks and the 9-month follow-up, 17 additional infants withdrew from the randomized trial. In the LCPUFA-formula group, there were 2 additional deaths (both while the infant was still receiving the trial diet). There were also 3 deaths in the breastfed group. All other withdrawn infants were lost to follow-up. Extra efforts were made to ensure completeness of the 18-month follow-up, and 5 infants who were not seen at 9 months attended for assessment at 18 months. More than 85% of visits at 9 months took place within 3 weeks of schedule, and >85% of 18-month visits were within 4 weeks of schedule. The follow-up rate for the randomized trial to 18 months was 84% (84 of 100) for control infants and 77% (74 of 96) for the LCPUFA-supplemented group (81% [74 of 91] for survivors).
There were no significant differences between infants seen or not seen at the 18-month follow-up (birth weight: 1318 g (SD: 271 g) versus 1378 g (266 g); gestation: 30.3 (2.0) versus 30.4 (2.2) weeks; 50% male in each group).
Baseline characteristics of the randomized diet groups are shown in Table 2. The groups were well matched for birth weight, gestation, gender, and the proportion of infants with birth weight below the 10th centile. The duration of hospital stay was similar in the 2 groups (38 [SD: 18] versus 36  days for control and LCPUFA-supplemented infants, respectively).
Primary Efficacy Outcome
All surviving infants were invited for follow-up, irrespective of whether they had completed a minimum of 3 weeks on the trial diet. At 18 months, scores obtained from the Bayley Scales of Infant Development did not differ significantly between groups, although infants from the LCPUFA group had a 2.6-point advantage in Bayley MDI and a 2-point advantage in PDI (Table 3).
Secondary Efficacy Outcomes
At 9 months of age, overall developmental scores and subscale scores differed little between randomized groups (Table 3). There were no significant differences between randomized groups in the proportion of infants considered by the examiner to have a possible or obvious neurologic deficit at 9 or 18 months.
Discharge weight was lower in infants who were fed the LCPUFA-supplemented formula (difference: 99 g; 95% confidence interval [CI]: 0.7–197; P = .047). However, discharge weight SD scores were not significantly different (difference: 0.24; 95% CI: −0.01–0.48; P = .06), and the change in SD score between birth and discharge did not differ between groups (Table 4).
At 9 months’ corrected age, infants who had received LCPUFA-supplemented formula were lighter (by 310 g [95% CI: −6–620 g]) and shorter (by 0.86 cm [−0.06–1.78]) than control infants. At 18 months, infants previously fed the LCPUFA-supplemented formula were significantly lighter (by 370 g [95% CI: 12–735]; P = .04) and shorter (by 1.5 cm [0.5–2.4]; P = .004) than control infants. The difference in weight and length between formula groups was present in both boys and girls (in boys, weight difference: 211 g [95% CI: −337–759); length difference: 1.05 cm [−0.55–2.64]; in girls, weight difference: 438 g [−31–907]; length difference: 1.70 cm [0.48–2.91]). However, there was no interaction between gender and diet on weight or length at 9 or 18 months. To adjust for the slight excess of girls in the LCPUFA-supplemented group, we expressed weight and length scores as SD scores. The length deficit at 18 months for LCPUFA-supplemented infants remained significant (0.44 SD; 95% CI: 0.08–0.8), whereas the difference in weight was reduced below the 5% significance level (0.33 SD; 95% CI: −0.02–0.69). Differences in weight and length at 18 months postterm remained after adjusting for parental smoking, social class, and level of maternal education. There were no differences in HC between randomized groups at either age.
We noted that the growth differences between randomized groups were greater in one center than the other (center 1: weight at 18 months 10.25 kg [SD: 1.05] control vs 10.11 kg (1.29) LCPUFA-supplemented [P = .6]; length at 18 months: 80.8 cm [3.0] vs 79.7 cm [3.2] [P = .09]; center 2: weight at 18 months: 10.37 kg [1.17] vs 9.70 kg [1.01] [P = .01]; length at 18 months: 80.8 cm [3.4] vs 78.9 cm [2.7] [P = .01]). However, there was no significant interaction between center and diet group on growth outcomes, and data from the 2 centers were therefore combined as planned in the analyses.
Adverse clinical events are recorded in Table 5. Four deaths occurred during the trial, all in infants who were randomized to the LCPUFA-supplemented formula (P = .06). However, 1 infant from the LCPUFA-supplemented formula group died from NEC on day 9, having received only 19 mL of trial formula. Excluding this infant gives a death rate of 0% versus 3.2% (P = .11). The 3 remaining infants who died had received significant volumes of LCPUFA-supplemented formula. All were extremely low birth weight (birth weights 990 g, 740 g, and 670 g), all had chronic lung disease requiring prolonged ventilation, and all were late deaths (46, 73, and 135 days). None of the deaths were believed by the attending neonatologist to be related to diet, and on close inspection of the clinical data it seemed that these infants’ problems were established before they were consuming significant volumes of enteral feeds.
The incidence of bacteriologically confirmed systemic infection was similar in the 2 groups (5% in LCPUFA-supplemented vs 7% in controls; P = .77). The incidence of confirmed NEC was 5.3% (5 of 95) in LCPUFA-supplemented infants versus 2% (2 of 100; P = .11) in controls. This figure, however, includes 2 infants who received only 13 mL and 19 mL of LCPUFA-supplemented formula before developing NEC and another who consumed no control formula before developing NEC. The volume of trial formula consumed by the remaining infants before developing NEC was 79, 93, and 2427 mL for LCPUFA-supplemented infants and 885 mL for the control infant.
The incidence of intraventricular hemorrhage, periventricular leukomalacia, patent ductus arteriosus requiring treatment, retinopathy of prematurity, and pulmonary hemorrhage did not differ between groups. There were no significant differences in the requirement for or duration of ventilation.
There were no significant differences between formula groups in the time taken to reach full enteral feeds (10.7 [SD: 6.3] vs 11.7 [8.6] days for control and LCPUFA-supplemented infants, respectively) or the number of days on which abdominal distension, soft stools, or nappy rash were reported by nursing staff. The mean number of stools per day was statistically significantly higher in the control group (median [25th, 75th centiles] 2.12 [1.65, 2.61] vs 1.96 [1.51, 2.48]; P = .04), and hard stools were reported in a greater proportion of these infants (37% vs 23%; P = .04).
Clinical Events at Follow-up (Beyond Period on Trial Diets)
Both formula groups had a similar incidence of upper respiratory tract infections, lower respiratory tract infections requiring antibiotics (“chest infections”), general practitioner visits, hospital admissions, and outpatient visits between discharge and the 18-month follow-up. The prevalence of eczema, wheeze, and asthma and the number of courses of antibiotics received between hospital discharge and 18 months also were similar in the 2 groups. Reported stool frequency and consistency at 9 and 18 months of age were not significantly different in infants previously fed control versus LCPUFA-supplemented formula.
Effect of Gestation
There was a statistically significant interaction between gestation and diet on Bayley PDI at 18 months of age (P = .05). To explore a possible influence of maturity on the response to LCPUFA supplementation, we therefore poststratified the cohort by gestation using the median value, 30 weeks, as a cutoff. Infants who had a gestational age <30 weeks and received the LCPUFA-supplemented formula had a Bayley MDI 4.5 points higher than infants who were fed the control formula (P = .2); for Bayley PDI, the advantage was 5.8 points (P = .1). In contrast, there were no differences in Bayley developmental indices between LCPUFA-supplemented and control infants with a gestational age >30 weeks (Table 6).
Effect of Volume of Trial Diet Consumed
The mean volume of trial formula consumed during the study was not significantly different in control and LCPUFA-supplemented infants (7389 [SD: 4239] vs 6375  mL). Median (25th, 75th centiles) for intake of LCPUFA-supplemented formula was 5902 (4354, 8054) mL. Exploratory analyses were performed to examine the principal outcomes in relation to both the duration and volume of trial formula consumed by the infant. There was no significant interaction between formula and duration or volume of formula consumed on later outcome and no evidence of a greater difference in any outcome between randomized diet groups in infants who had consumed the greatest volumes of formula.
Comparison of Breastfed Reference Group With Formula Groups
Breastfed infants had similar birth weight and gestation to formula-fed infants but were less likely to have been small for gestational age at birth. They had significantly lower weight gain while in the hospital than both formula groups (Table 4).
At 9 months of age, there were no significant differences in weight, length, or HC between breastfed infants and either formula group; measurements for breastfed infants were similar to those for control infants. At 18 months, breastfed infants were significantly heavier and longer than LCPUFA-supplemented infants and similar to the controls; these results were not altered by converting measurements to SD scores. These differences remained after adjusting for social class, level of maternal education, and parental smoking. There were no significant differences in HC between diet groups.
Developmental quotients at 9 months were significantly higher in breastfed infants than in both formula groups on the adaptive, gross motor, fine motor, and personal-social subscales and for overall quotient; lowest scores were seen in control infants with intermediate scores in LCPUFA-supplemented infants (Table 3). At 18 months of age, breastfed infants had significantly higher Bayley MDI and PDI scores than both formula groups (Table 3). The advantage in developmental scores for breastfed infants at both ages persisted after adjusting for potentially confounding variables (social class, level of maternal education, birth order, and marital status). Thus, for Bayley MDI, the adjusted advantage for the breast milk-fed group was 12.9 points (95% CI: 6.2–19.5) compared with the controls and 9.3 points (2.8–15.8) compared with the LCPUFA group, and, for Bayley PDI, the advantage was 8.7 points (2.2–15.2) compared with controls and 7.1 points (0.6–13.5) compared with LCPUFA-supplemented infants.
Compared with both formula groups, breastfed infants had a lower incidence of hematologically confirmed infection (6.8% vs 17.9%; P = .02), although the difference in more stringently (bacteriologically) confirmed infection was not significant (2.3% vs 6.1%; P = .2). A total of 3.6% of formula-fed infants developed confirmed NEC compared with none of the breastfed infants (P = .1). There were no significant differences in rates of illness or atopy or in antibiotic usage at 9 or 18 months between breastfed and formula-fed infants.
Use of an LCPUFA-supplemented formula in preterm infants did not produce a significant benefit for neurodevelopment at 9 or 18 months of age. Our study was powered to detect a difference of 4 points (0.4 SD) in Bayley developmental indices between randomized groups; such a difference is plausible and is smaller than that found between randomized diet groups in our previous studies of preterm infants18 and between infants who received banked breast milk compared with preterm infant formula.19 We cannot, however, exclude a smaller overall effect of LCPUFA supplementation on development. LCPUFA-supplemented infants in our study had Bayley MDI scores that were 2.6 points higher than control infants (approximately 0.25 SD); a much larger study, involving approximately 256 infants per randomized group, would be required to detect a benefit of this magnitude as significant. In addition, there was some evidence for a greater benefit of LCPUFA-supplemented formula on Bayley developmental indices at 18 months in the most immature infants. Those who were on the LCPUFA-supplemented formula and were <30 weeks’ gestation had MDI and PDI scores 4.5 and 5.8 points higher, respectively, than the controls. Approximately 60 to 70 such infants per randomized group would have been needed to test for a difference of this magnitude.
The LCPUFA-supplemented formula used in our study was manufactured in the mid-1990s and contained a lower concentration of DHA than that recently recommended by a workshop of investigators in the field,20 although a formula with even lower DHA content than ours was shown previously to produce clear biochemical effects in preterm infants.21 In addition, the period of supplementation in our study was shorter than that in some other studies. It is possible that these 2 factors may contribute to the relatively small observed effect of supplementation on neurodevelopmental outcome.
Developmental findings in other published trials of LCPUFA supplementation in preterm infants have been inconclusive, with studies showing both worse22–24 and better25,26 outcome in supplemented infants. However, failure to demonstrate a significant effect of LCPUFA supplementation on global tests of development during infancy does not exclude the possibility of more subtle effects, which might become apparent at a later age, when more detailed neurocognitive testing is possible. This emphasizes the need for long-term follow-up of infants enrolled in supplementation studies, currently being planned for our cohort.
A specific aim of our study was to investigate safety issues. Two early studies13,27 identified growth deficits persisting for up to 1 year of age, well beyond the period of LCPUFA supplementation. In both studies, infants had received marine oil as a source of n-3 fatty acids without an additional supply of n-6 fatty acids. They became relatively deficient in AA, and this was thought to explain the growth deficit. Another study, by Montalto and colleagues,28 also showed reduced growth to 6 months in male preterm infants fed a formula with n-3 LCPUFA.
Subsequently, 5 randomized trials of LCPUFA supplementation in preterm infants have found no effect on growth,26,29–32 and 1 study reported higher weights in supplemented infants compared with controls at 2 months but not at 4 months postterm.33 All 6 trials used a balanced addition of n-3 and n-6 fatty acids, leading to the conclusion in a recent review for the Cochrane Library10 that LCPUFA supplementation does not impair the growth of preterm infants. In this light, our finding of significant growth deficits at 18 months in LCPUFA-supplemented infants is surprising, particularly as growth rates during the period on the trial diets were not significantly different in supplemented and control infants. The disadvantage in weight at 18 months for LCPUFA-supplemented infants was reduced when measurements were expressed as SD scores to correct for minor imbalances in gender or age at follow-up between groups. However, the length deficit remained, amounting to 0.44 SD, or 9%, of the population variance assuming a normal distribution. Although not necessarily important for an individual, this magnitude of effect would be more relevant in population terms.
It is interesting that the differences in weight and length between randomized groups were greater at 18 months than at 9 months postterm, suggesting progressive amplification of the growth effect with age. Such longitudinal amplification effects on growth have also been seen in rats after more general experimental alterations in nutrition in the newborn period.34 With the exception of 1 study,32 follow-up was shorter in the previous LCPUFA studies, in which no differences in growth were identified (up to term, 6 weeks of age, or 4 months postterm), whereas in the Carlson studies, infants were followed up to 1 year postterm. Thus, it is possible that effects of LCPUFA supplementation on growth are delayed and therefore more likely to be seen in studies with longer follow-up. Moreover, the precise nature, dosage, and duration of LCPUFA supplementation used might influence any effects on growth. Clearly, additional work is needed to exclude the possibility that this is a chance finding.
There were 4 deaths in the randomized trial, all in LCPUFA-supplemented infants. One infant died after receiving only 19 mL. The remaining 3 infants who died were late deaths. All were of extremely low birth weight and had chronic lung disease requiring prolonged ventilation. Inspection of their clinical progress suggested that their problems were well established before the establishment of enteral feeds, and in no case was diet believed by the attending neonatologist to be involved in the infant’s death.
In LCPUFA-supplemented infants, there were 5 cases (5.2%) of NEC compared with 2 (2%) in controls. This nonsignificant difference was reduced further after excluding 2 supplemented infants who received <20 mL of formula and 1 control infant who received no formula at all before developing NEC. A similar pattern was reported by Carlson27: 9 of 47 LCPUFA-supplemented infants developed NEC compared with 3 of 47 controls (P = .06). In a more recent trial35 using a different preterm formula with LCPUFA derived from egg phospholipids, supplemented infants had significantly less NEC (2.9% vs 17.6% in controls). These findings emphasize the need for additional investigation of safety issues, both in new, larger randomized interventions targeted to explore safety as well as efficacy and by meta-analysis.
Uauy29 reported that preterm infants who were fed a formula supplemented with fish oil had significantly longer bleeding times than infants who were fed control formula, although both were thought to be in the normal range. We found no evidence of an increase in hemorrhagic clinical events in LCPUFA-supplemented infants.
As AA and DHA are present in breast milk, they might be regarded as inherently safe. However, available sources of AA and DHA are present in complex oils, some of which contain “unphysiological” lipids (not found in breast milk). Given the possible bioactivity of exogenous LCPUFA and the differences in dosage and duration of LCPUFA supplementation used in various studies, it is not surprising that consistent safety and, indeed, efficacy findings have not emerged. Although future meta-analysis will be important in increasing the power to detect group differences, there must be some caution in such analyses in combining studies performed with different preparations and dosing conditions. This is well illustrated by the different cognitive development findings reported recently in a study that used LCPUFA from 2 different sources.26 Ultimately, reassurance on safety and proof of efficacy may be best obtained from detailed study of individual formula products.
Comparison of Formula-Fed Infants With Breastfed Reference Group
Breastfed infants showed, as expected,36 significantly lower weight gains than formula-fed infants during the period in the hospital. They also had significantly fewer hematologically confirmed infections and a trend toward a decrease in both more stringently confirmed infections and NEC compared with both formula groups. These data add to the increasing evidence for a protective effect of breast milk in neonatal intensive care.
Breastfed infants also showed the expected advantages in developmental quotient at 9 months and in Bayley MDI and PDI at 18 months compared with both formula groups; differences were greatest between the control group and the breastfed group and remained after adjusting for potential socioeconomic and educational confounding variables. The hypothesis that the advantage in cognitive development for breastfed versus formula-fed infants could be accounted for by LCPUFA present in breast milk was not supported by our data as LCPUFA-supplemented infants did not attain the same cognitive scores as breastfed infants, although this will continue to be tested, using additional outcomes, at follow-up of this cohort.
It is interesting that breast milk-fed infants were significantly heavier and longer than LCPUFA-supplemented infants at the 18-month follow-up but similar in size to those fed the control formula. We did not collect data on parental size and could not, therefore, adjust for any genetic growth potential differences in the breast versus formula groups. Nevertheless, our findings provide additional evidence that the LCPUFA-supplemented group had reduced growth up to 18 months.
In this large study, we found no significant beneficial effect of LCPUFA supplementation during the neonatal period in preterm infants on their developmental outcome in the first 18 months. However, given that there was a nonsignificant advantage in the developmental scores at 18 months in the LCPUFA group, most marked in infants <30 weeks’ gestation, additional study is needed to test whether there is indeed an underlying benefit or whether this difference was attributable to chance. It is particularly important that this cohort be followed up at an age when more detailed and specific testing is possible, and this is planned.
Our study raised a potential safety issue. Reduced linear growth in the LCPUFA group (compared with both randomized controls and a breast milk-fed reference group) became amplified over time and was seen 18 months after dietary randomization ceased. The higher death rate in the LCPUFA group seemed unlikely to be causal after exploratory analyses, and the higher incidence of chronic lung disease and NEC in the LCPUFA-supplemented group was not statistically significant. Our results suggest, however, that residual efficacy and safety issues in the LCPUFA supplementation of formula for preterm infants require additional testing.
This study was supported by a grant from Numico Research BV (Wageningen, The Netherlands), who also provided the infant formulas.
We thank the research staff who collected data in the study (Julie Owen, Geraldine McHugh, Mary Quinn, Dawn Rodd, Emma Sutton, and Catherine Leeson-Payne) and the parents who allowed their infants to participate. We also thank Numico for collaboration, contributory funding, and supply of the trial formulas.
- Received June 5, 2001.
- Accepted December 18, 2001.
- Reprint requests to (M.S.F.) MRC Childhood Nutrition Research Center, Institute of Child Health, 30 Guilford St, London WC1N 1EH. Email:
- ↵Makrides M, Neumann MA, Byard RW, et al. Fatty acid composition of the brain, retina and erythrocytes in breast- and formula-fed infants. Am J Clin Nutr.1994;60 :180– 194
- ↵Simmer K. Longchain polyunsaturated fatty acid supplementation in infants born at term (Cochrane Review). Cochrane Database Syst Rev.2000;3 :CD000376
- ↵Morley R, Cole TJ, Powell R, Lucas PJ. Mother’s choice to provide breast milk and developmental outcome. Arch Dis Child.1988;63 :1382– 1385
- ↵Bayley N. Manual for the Bayley Scales of Infant Development. 2nd ed. San Antonio, TX: The Psychological Corporation; 1993
- ↵Freeman JV, Cole TJ, Chinn S, Jones PRM, White EM, Preece MA. Cross-sectional stature and weight reference curves for the UK, 1990. Arch Dis Child.1995;73 :17– 24
- ↵Carlson SE, Werkman SH, Peeples JM, Cooke RJ, Tolley EA. Arachidonic acid status correlates with first year growth in premature infants. Proc Natl Acad Sci U S A.1993;90 :1073– 1077
- ↵Carlson SE. Long-chain polyunsaturated fatty acid supplementation of preterm infants. In: Dobbing J, ed. Developing Brain and Behavior: The Role of Lipids in Infant Formula. San Diego, CA: Academic Press Ltd (Ross Pediatrics); 1997:66
- ↵O’Connor D, Hall R, Adamkin D, et al. Growth and development in preterm infants fed long-chain polyunsaturated fatty acids: a prospective, randomized controlled trial. Pediatrics.2001;108 :359– 371
- ↵Carlson SE, Werkman SH, Tolley EA. Effect of long-chain n-3 fatty acid supplementation on visual acuity and growth of preterm infants with and without bronchopulmonary dysplasia. Am J Clin Nutr.1996;63 :687– 697
- ↵Hansen J, Schape D, et al. Docosahexanoic acid plus arachidonic acid enhance preterm infant growth. Essential fatty acids and eicosanoids (Edinburgh) 1997:T16
- ↵Lucas A, Gore SM, Cole TJ, et al. Multicentre trial on feeding low birthweight infants: effects of diet on early growth. Arch Dis Child.1984;59 :722– 730
- Copyright © 2002 by the American Academy of Pediatrics