Human milk contains long-chain polyunsaturated fatty acids (LCPUFA), most notably n-3 docosahexaenoic acid (DHA) and n-6 arachidonic acid (AA). Advantages in cognitive development of breast-fed in comparison with formula-fed infants have been attributed to the lack of these LCPUFAs in the past standard commercially available formulas
Short-term beneficial effects of LCPUFAs such as DHA and AA on cognitive and visual development of the offspring were shown in several studies
, but the long-term effect of these LCPUFAs on infant mental and psychomotor development at the age of 3 years, for example, is controversial (2–4) . Guided by the short-term beneficial effects of LCPUFAs, Koletzko et al have recommended the use of formulas with a proportion of DHA between 0.2% and 0.5% of fatty acids and amounts of AA being at least equal to those of DHA (5,6) . (7)
In 1- to 2-year-old children, in particular, contradictory findings have been reported
. It is unclear whether these differences may be attributed to whether term or preterm infants had been included in the study, to the LCPUFA amount, composition and source, population studied, and outcome measures used (8–21) . (22,23)
Systematic reviews on term and preterm children concluded that there is little evidence for benefits to LCPUFA-supplemented children in
neurodevelopment, but small effects of LCPUFA supplementation may have escaped detection because of lack of power, the impossibility to adjust for confounders, and to perform subgroup analyses in the conventional meta-analysis approach . (24–26)
These limitations may be overcome by an individual patient data (IPD) meta-analysis on the raw data of relevant original studies. Individual patient data meta-analyses are regarded as providing “the least biased and most reliable means” to combine results from different studies
. However, this advantage may be challenged if data are not available from all relevant studies and if the assessment of relevant covariates differs substantially between the included studies. (27)
We performed an IPD meta-analysis based on 4 independent large clinical trials
on LCPUFA supplementation in term or preterm infants, assessing (8–11) neurodevelopment at 18 months using Bayley Scales of Infant Development II (BSID II) . The data were provided in the framework of the Early Nutrition Programming Project, a large European Union-funded research consortium exploring long-term consequences of nutrition during pregnancy and infancy on development and health (28) . Unfortunately, we did not have access to data from 2 other relevant studies (29) . (13,18)
The question addressed was whether there is a global effect of LCPUFA supplementation on
neurodevelopment, or a potential effect confined to subgroups of preterm infants, very-low-birth-weight infants (VLBW), boys, girls, or regarding specific composition or sources of LCPUFA. MATERIALS AND METHODS
Search for Other Relevant Studies
Because this IPD meta-analysis started with the data provided within the Early Nutrition Programming Project consortium, an effort was made to include all of the relevant data. We searched in MEDLINE (November 2007) for other randomised trials examining the effect of LCPUFA supplementation on
neurodevelopment of infants measured by BSID II at 18 months of age. Search items were “Bayley” or “development” and “LCPUFA.” From 8 studies with adequate setting , only 2 proved to be relevant (12–19) , because the others had assessed mental developmental index (MDI) and psychomotor development index (PDI) at time points other than 18 months of age; moreover, 2 of those had used BSID I ( (13,18) Table 1). Attempts to access the data from the relevant studies were not successful. We were not aware of other unpublished studies. TABLE 1:
Other randomised studies on the effect of LCPUFA supplementation on BSID indexed in MEDLINE
The meta-analysis was based on individual patient data from randomised clinical trials from Groningen
, Leicester, Nottingham (8) , and Glasgow (9,10) . Two of the British studies involved preterm infants (11) . Here, the Leicester/Nottingham term and preterm studies will be called Leicester 1 and Leicester 2, respectively. The Leicester 2 children had a gestational age of <37 weeks and a birth weight <1750 g, the infants from the Glasgow study had a gestational age <35 weeks and a birth weight ≤2000 g. (10,11)
In all of the trials, infants were randomised to receive either infant formulas with additional LCPUFA (LF) or a control unsupplemented infant formula (CF) in a double-blind design. The infants were not randomised to other interventions. The trial formulas were supplemented with n-3 DHA and n-6 AA and in the 2 preterm trials additionally with n-6 γ-linolenic acid (GLA). The control formulas were virtually free of DHA, AA, and GLA. The studies differed in formula composition, duration of formula feeding (
Table 1), and formula sources: The LCPUFAs in the Groningen trial came from egg, fish oil (DHA), and single cell oil (AA), whereas the LCPUFAs in the UK trials contained mixtures of egg lipid/phospholipid, fish oil, and borage oil. The DHA content of formulas ranged from 0.17% to 0.5% of fatty acids, with varying DHA/AA ratios ( Table 2). All 4 studies measured BSID II at 18 months corrected age as outcome variable. The Groningen trial used the Dutch version (BSID-II-NL). TABLE 2:
Formula compositions of LCPUFA-supplemented formulas in grams per 100 g fat and duration of formula feeding in the 4 trials
In the term studies from Groningen and Leicester 1, data on MDI and PDI were available in 279 (132 LF + 147 CF) and 250 (125 LF +125 CF) cases, respectively. The preterm trial from Leicester 2 contained 147 (68 LF +79 CF) children, and the Glasgow study contained 194 (103 LF +91 CF). Therefore, the meta-analysis included 529 (257 LF +272 CF) term infants, 341 (171 LF +170 CF) preterm infants, and 870 (428 LF +442 CF) infants in total, with the numbers of LF and CF infants almost equal.
Preterm delivery as defined in the studies
Because different definitions for prematurity had been applied in the respective individual trials, we also generated the subgroup VLBW with children with birth weight of <1500 g (n = 175 = 95 LF + 80 CF), all of whom were born before the 35th week of gestation.
Because the level of DHA in the Leicester 2 trial was relatively low (
Table 2), we investigated the subgroup “higher DHA levels,” including only the other 3 trials, which had used formulas with the recommended level of DHA between 0.2 and 0.5 . (7) Equivalently, we examined data from children supplemented with high AA levels (AA ≥ 0.2), leaving out the Glasgow study.
For the last subgroup we considered all 3 UK trials only because their LCPUFA sources in the formula differed from those in the Groningen study.
For all of the infants as well as for the 8 subgroups defined above, mean differences in Bayley MDI and PDI were calculated, adjusted for confounders, and tested by Student
t test (2-sided hypothesis). This was done by applying multivariable linear models with BSIDs as outcome variables, LCPUFA supplementation as explanatory variable, and sex, gestational age, birth weight, maternal age, and maternal smoking during the third trimester as confounders. A significant mean difference between LF and CF groups in MDI and PDI was set at an α level of 0.05.
Although information on sociodemographics reported as maternal education (basic school at best vs certificate qualifying for university) was provided in all of the studies, we did not include this potential confounder in the main analyses because of missing data in 42% of the mothers in the Leicester 2 trial (7% in total); however, maternal education was considered in additional sensitivity analyses. Other sensitivity analyses looked for potential differences of the effects in boys and girls in subgroups 1 to 6.
We also examined a potential dose-response effect of DHA on MDI and PDI by calculating confounder-adjusted linear models replacing the explanatory variable LCPUFA supplementation with the variable amount of DHA given in the respective study (set to “0” for the CF group). From these analyses we excluded the data from the Groningen trial because the Bayley scores in this trial were higher than in the UK trials.
Power of the Study
For our power calculations, we assumed an α level of 0.05, a desired statistical power of 0.8, standard deviations of MDI and PDI of 15, and equal sizes of the number of LF and CF infants in every subgroup.
Data of 870 children permitted detection of a difference of 2.9 points in MDI or PDI, whereas 341 preterm children would be enough to find a score difference of 4.6. The subgroup of 175 VLBW infants would still allow detection of a difference of 6.4 score points in MDI or PDI.
As expected, the term children showed higher birth weight and gestational age and a smaller percentage of mothers smoking in the third trimester compared with the preterm children, whereas there were no significant differences in any of these variables for LCPUFA versus control infants (
Table 3). Maternal age was comparable across all groups. However, the proportion of mothers with high levels of education was significantly higher in children randomised to LCPUFA supplementation. TABLE 3:
Means (standard errors) and numbers of cases (%) of maternal and infant characteristics of term, preterm, and all children of all 4 included studies
Crude analyses are presented in
Table 4. There were no significant differences between LF and CF groups in MDI and PDI overall and in any subgroups. Also, after adjustment for sex, gestational age, birth weight, maternal age, and maternal smoking, no significant mean differences were found between LCPUFA supplemented and control groups in MDI (difference −0.8 [95% confidence interval −2.8 to 1.2]) and PDI (−1.0 [−2.7 to 0.7]). Furthermore, there were no significant findings in any of the defined subgroups ( Table 5). TABLE 4:
Mean values of MDI and PDI and unadjusted mean differences of LCPUFA vs control children in subgroups with term, preterm, and VLBW infants, high DHA and AA levels, children from the UK trials (receiving LCPUFAs from sources of egg lipid/phospholipid, fish oil and borage oil), boys, girls, and all infants
Mean differences in MDI and PDI of LCPUFA vs control children in subgroups with term, preterm; and VLBW infants, high DHA and AA levels, children from the UK trials (receiving LCPUFAs from sources of egg lipid/phospholipid, fish oil, and borage oil), boys, girls, and all infants
Sensitivity analyses revealed identical findings in boys and girls overall, in all subcategories, and with inclusion of higher maternal education as a potential confounder (data not shown). The variable “amount of DHA” was no significant predictor in the dose response analyses.
In contrast to classical meta-analyses based on aggregated data from different studies, this IPD meta-analysis on a large sample size allowed adjustment for confounders and to perform subgroup analyses. The absence of any detectable benefit or disadvantage in
neurodevelopment assessed with BSID at the age of 18 months for all of the children or in any subgroup therefore provides evidence against beneficial effects of LCPUFA supplementation on BSID at 18 months under the conditions of the trials included here. The strength of the evidence depends substantially on the sample size of the data included and on the results of studies for which data were not available. In any case, our results do not exclude potential benefits under specific conditions, such as the reported improvement of BSID at 18 months in male preterm infants provided with a higher dosage of DHA of 0.5% of fatty acids . (11)
The strength of our meta-analysis is its sample size with sufficient power to detect meaningful differences in Bayley scores, not only in the entire sample but also in subgroups. A difference of at least 5 points in the Bayley scores was considered as clinically relevant, taking account of the standard errors of measurement of MDI (4.27) and PDI (5.69) at 18 months given in the test manual
. The sizes of the total sample and of the subgroups are large enough to find such a relevant difference, except for the subgroup of VLBW infants with a detectable difference of slightly above 5 points. (28)
The only relevant data of term children not included in the IPD meta-analysis were those from the Birch et al
study, which had shown beneficial effects of LCPUFA supplementation on (13) neurodevelopment; however, this trial included only 56 children. In our meta-analysis, supplemented term infants tended to have lower Bayley scores than controls. Therefore, inclusion of the Birch et al data would be unlikely to change the results of our meta-analysis substantially (13) . (30)
The nonavailability of the data from the study of Clandinin et al
, however, was a bigger concern. In the mentioned study, a control group (n = 54) was compared with 2 groups supplemented with different LCPUFAs (N = 104). All of the children included were born preterm, and almost all of them were born with VLBW (≤1500 g). In the control group, mean values of 77.2 in MDI and 83.0 in PDI were observed, in contrast to 85.1 (MDI) and 90.7 (PDI) in the combined group of supplemented children (18) . Based on these results and the unadjusted mean differences from our study, weighted mean differences (95% confidence intervals) of preterm children were 3.1 (0.3–6.0) in MDI and 2.1 (−0.8 to 5.1) in PDI and were therefore of no clinical relevance (as defined above). In respect of VLBW infants only, differences of 4.2 (0.5–7.9) and 4.7 (0.8–8.7) were detected in MDI and PDI, respectively. Although these appraisements did not consider possible confounding effects, they indicate that inclusion of the Clandinin et al (26) data would probably not have led to differing conclusions in other subgroups than the VLBW group. (18)
These conclusions are in accordance with those from 2 recently published Cochrane reviews
, which were based on the studies included in our IPD meta-analysis (24,25) and those from Birch et al (8–11) and Clandinin et al (13) . These classical meta-analyses did not detect significant benefits in MDI and PDI of term and preterm children at 18 months of age. Another review on the effect of LCPUFA supplementation (18) on (26) neurodevelopment included 2 additional studies that had assessed BSID II earlier than 18 months . This meta-analysis found a significant increase of MDI by 3.4 points, but no significant change in PDI. Again, it is disputable whether the observed change in MDI is of clinical relevance. Furthermore, the observed effects disappeared when the results of 2 studies with BSID I as outcome (15,19) were additionally considered. (12,16)
Other possible weaknesses of this study may arise from limitations regarding the data provided. One of the trials lacked information on maternal education in about half of the mothers; however, including higher maternal education into multivariable analyses as an additional covariate where available had no considerable effects on the results. Furthermore, systematically higher MDI and PDI were reported in the Groningen trial. This may be because of different reference norms in the original and translated version of the Bayley test, but is unlikely to interfere with supplementation effects because of almost equal numbers of LF and CF groups in the Groningen trial.
Other studies detected benefits of LCPUFA supplementation on other neurological outcomes, such as general movements
or visual acuity (2) in children younger than 12 months during infancy. A previous trial in 18-month-old children detected effects of prenatal fatty acid status on neurologic optimality scores, but none on BSID (3,4) . The reason may be that BSIDs, the most frequently used scores to assess (21) neurodevelopment of infants , are a less subtle measurement than neurologic optimality scores and may have limitations in detecting differences in “excellence.” (31)
Measurements based on BSID at 18 months may miss subtle differences related to LCPUFA supplementation and may miss effects manifesting at later ages. For example, in a large double-blind randomised trial providing 200 mg DHA per day or placebo to breast-feeding women, Bayley scores at 18 months were not affected, but the supplemented group showed significantly better psychomotor development scores at 2½ years
. Similarly, randomised clinical trials providing oils with considerable amounts of DHA to women beginning in pregnancy found improved cognitive development at later ages of 2½ years (32) and at 4 years (33) , whereas another study showed no improvement of visual acuity and intelligence quotient for LCPUFA-supplemented children at 3 years of age (34) . A dose-response effect between maternal n-3 LCPUFA intake from seafood and children's verbal intelligence quotient at the age of 8 years, after adjustment for 28 confounding factors, was reported in the Avon Longitudinal Study of Parents and Children (23) . (35)
This IPD meta-analysis provides substantial evidence that there are no clinically relevant effects of LCPUFA supplementation on
neurodevelopment as assessed by BSID at 18 months in almost all subgroups except for VLBW children. For these, additional inclusion of data from all of the relevant studies in this field may have resulted in the detection of relevant beneficial effects. Possible effects of LCPUFA on other outcomes, including outcomes at later ages, deserve further study. REFERENCES
1. Agostoni C. LC-PUFA content in human milk: is it always optimal? Acta Paediatr 2005; 94:1532–1534.
2. Bouwstra H, Dijck-Brouwer DA, Wildeman JA,
. Long-chain polyunsaturated fatty acids have a positive effect on the quality of general movements of healthy term infants. Am J Clin Nutr 2003; 78:313–318.
3. Lauritzen L, Jørgensen MH, Mikkelsen TB,
. Maternal fish oil supplementation in lactation: effect on visual acuity and n-3 fatty acid content of infant erythrocytes. Lipids 2004; 39:195–206.
4. Makrides M, Neumann M, Simmer K,
. Are long-chain polyunsaturated fatty acids essential nutrients in infancy? Lancet 1995; 345:1463–1468.
5. Hadders-Algra M. The role of long-chain polyunsaturated fatty acids (LCPUFA) in growth and development. Adv Exp Med Biol 2005; 569:80–94.
6. Innis SM. Fatty acids and early human development. Early Hum Dev 2007; 83:761–766.
7. Koletzko B, Lien E, Agostoni C,
. The roles of long-chain polyunsaturated fatty acids in pregnancy, lactation and infancy: review of current knowledge and consensus recommendations. J Perinat Med 2008; 36:5–14.
8. Bouwstra H, Dijck-Brouwer DAJ, Boehm G,
. Long-chain polyunsaturated fatty acids and neurological development outcome at 18 months in healthy term infants. Acta Paediatr 2004; 94:26–32.
9. Lucas A, Stafford M, Morley R,
. Efficacy and safety of long-chain polyunsaturated fatty acid supplementation of infant-formula milk: a randomised trial. Lancet 1999; 354:1948–1954.
10. Fewtrell MS, Morley R, Abbott RA,
. Double-blind, randomized trial of long-chain polyunsaturated fatty acid supplementation in formula fed to preterm infants. Pediatrics 2002; 110(1 Pt 1):73–82.
11. Fewtrell MS, Abott RA, Kennedy K,
. Randomized, double-blind trial of long-chain polyunsaturated fatty acid supplementation with fish oil and borage oil in preterm infants. J Pediatr 2004; 144:471–479.
12. Carlson SE. Lipid requirements of very-low-birth-weight infants for optimal growth and development. In: Dobbing J, ed.
Lipids, Learning, and the Brain: Fats in Infant Formulas.
Columbus, OH: Ross Laboratories; 1993:188–214.
13. Birch EE, Garfield S, Hoffman DR,
. A randomized controlled trial of early dietary supply of long-chain polyunsaturated fatty acids and mental development in term infants. Dev Med Child Neurol 2000; 42:174–181.
14. Makrides M, Neumann M, Simmer K,
. A critical appraisal of the role of dietary long-chain polyunsaturated fatty acids on neural indices of term infants: a randomized, controlled trial. Pediatrics 2000; 105(1 Pt 1):32–38.
15. O'Connor DL, Hall R, Adamkin D,
. Growth and development in preterm infants fed long-chain polyunsaturated fatty acids: a prospective, randomized controlled trial. Pediatrics 2001; 108:359–371.
16. Van Wezel-Meijler G, van der Knaap MS, Huisman J,
. Dietary supplementation of long-chain polyunsaturated fatty acids in preterm infants: effects on cerebral maturation. Acta Paediatr 2002; 91:942–950.
17. Agostoni C, Verduci E, Massetto N,
. Plasma long-chain polyunsaturated fatty acids and
through the first 12 months of life in phenylketonuria. Dev Med Child Neurol 2003; 45:257–261.
18. Clandinin MT, Van Aerde JE, Merkel KL,
. Growth and development of preterm infants fed infant formulas containing docosahexaenoic acid and arachidonic acid. J Pediatr 2005; 146:461–468.
19. Fang PC, Kuo HK, Huang CB,
. The effect of supplementation of docosahexaenoic acid and arachidonic acid on visual acuity and
in larger preterm infants. Chang Gung Med J 2005; 28:708–715.
20. Agostoni C, Trojan S, Bellù R,
. Developmental quotient at 24 months and fatty acid composition of diet in early infancy: a follow up study. Arch Dis Child 1997; 76:421–424.
21. Bouwstra H, Dijck-Brouwer J, Decsi T,
. Neurologic condition of healthy term infants at 18 months: positive association with venous umbilical DHA status and negative association with umbilical trans-fatty acids. Pediatr Res 2006; 60:334–339.
22. Fewtrell MS. Long-chain polyunsaturated fatty acids in early life: effects on multiple health outcomes. A critical review of current status, gaps and knowledge. Nestle Nutrition Workshop Ser Pediatr Program 2006; 57:203–222.
23. Auestad N, Scott DT, Janowsky JS,
. Visual, cognitive, and language assessments at 39 months: a follow-up study of children fed formulas containing long-chain polyunsaturated fatty acids to 1 year of age. Pediatrics 2003; 112(3 Pt 1):e177–e183.
24. Simmer K, Patole SK, Rao SC. Longchain polyunsaturated fatty acid supplementation in infants born at term.
Cochrane Database Syst Rev
25. Simmer K, Schulzke SM, Patole S. Longchain polyunsaturated fatty acid supplementation in preterm infants.
Cochrane Database Syst Rev
26. Smithers LG, Gibson RA, McPhee A,
. Effect of long-chain polyunsaturated fatty acid supplementation of preterm infants on disease risk and
: a systematic review of randomized controlled trials. Am J Clin Nutr 2008; 87:912–920.
27. Stewart LA, Parmar MK. Meta-analysis of the literature or of individual patient data: is there a difference? Lancet 1993; 341:418–422.
28. Bayley N.
Bayley Scales of Infant Development
. 2nd ed. San Antonio, TX: Psychological Corporation; 1993.
29. Fewtrell MS. EARNEST Consortium session 6: infant nutrition: future research developments in Europe. EARNEST, The Early Nutrition Programming Project: early nutrition programming—long-term efficacy and safety trials and integrated epidemiological, genetic, animal, consumer and economic research. Proc Nutr Soc 2007; 66:435–441.
30. Egger M, Smith GD, Altman DG, eds.
Systematic Reviews in Health Care: Meta-Analysis in Context
. London: BMJ Books; 2001.
31. Johnson S, Marlow N, Wolke D,
. Validation of a parent report measure of cognitive development in very preterm infants. Dev Med Child Neurol 2004; 46:389–397.
32. Jensen CL, Voigt RG, Prager TC,
. Effects of maternal docosahexaenoic acid intake on visual function and
in breastfed term infants. Am J Clin Nutr 2005; 82:125–132.
33. Dunstan JA, Simmer K, Dixon G,
. Cognitive assessment of children at age 2(1/2) years after maternal fish oil supplementation in pregnancy: a randomised controlled trial. Arch Dis Child Fetal Neonatal Ed 2008; 93:F45–50.
34. Helland IB, Smith L, Saarem K,
. Maternal supplementation with very-long-chain n-3 fatty acids during pregnancy and lactation augments children's IQ at 4 years of age. Pediatrics 2003; 111:e39–e44.
35. Hibbeln JR, Davis JM, Steer C,
. Maternal seafood consumption in pregnancy and neurodevelopmental outcomes in childhood (ALSPAC study): an observational cohort study. Lancet 2007; 369:578–585.