Long-chain PUFA (LCPUFA) have become a major focus of attention in the field of infant nutrition and development. The Cochrane reviews of the group of Simmer(Reference Simmer, Patole and Rao1, Reference Simmer, Schulzke and Patole2) indicated that supplementation of formula with LCPUFA, in particular DHA (22 : 6n-3) and arachidonic acid (20 : 4n-6), in general does not affect motor, visual and cognitive outcome. However, when the age at which the child is assessed is taken into account the results are less straightforward. Outcome of LCPUFA-supplemented formula groups (especially DHA supplementation) is better in early infancy, but studies which assessed outcome between 6 and 24 months usually did not demonstrate differences between supplemented and non-supplemented groups(Reference Simmer, Patole and Rao1). The latter age period is, however, known for its insensitivity to reveal adverse effects of exposures during early ontogeny(Reference Hadders-Algra3). Thus, the absence of an effect between 6 and 24 months does not preclude an effect at a later age. Currently data on the effect of LCPUFA supplementation of formula on neurodevelopmental outcome at school age are lacking.
The present study aims at investigating whether consumption by term infants of formula containing LCPUFA for the first 2 months after birth improves the neurological condition of these children at 9 years of age. Neurological development is one of the domains in which the early beneficial effects of LCPUFA manifest themselves. The study is an extension of the original Groningen LCPUFA study, a randomised controlled trial on the effect of supplementation of formula with LCPUFA during the first 2 postnatal months. At the age of 9 years a standardised and age-specific neurological examination was carried out, which specifically assesses minor neurological dysfunction (MND). MND is considered a valid measure of neurological condition that shows a strong relationship with pre- and perinatal adversities(Reference Hadders-Algra3). Earlier, Lanting et al. (Reference Lanting, Fidler and Huisman4) reported that breast-feeding was associated with less MND at school age than formula-feeding. The primary outcome measure of the present study was the Neurological Optimality Score (NOS). The NOS uses the optimality concept to summarise neurological condition; it provides a sensitive measure of the child's overall neurological status. Secondary outcome measures were clinical neurological condition in terms of severity and type of MND. Two major forms of MND can be distinguished: simple and complex MND. Simple MND can be regarded as a typical but non-optimal form of brain function, whereas complex MND is the clinically relevant form of MND.
Based on the literature, we hypothesise that the LCPUFA-supplemented children perform better than the children who received standard formula and that breast-fed children perform better than the formula-fed children. A better performance will be reflected in a higher NOS, less and less severe MND, and a lower prevalence of specific neurological dysfunctions. The effect of sex on relationships between postnatal supplementation and neurological outcome was also determined, as previous studies suggested the possibility of an advantage of LCPUFA supplementation in males(Reference Hadders-Algra, Bouwstra and van Goor5).
Methods
Participants
Details of the study design have been described previously(Reference Bouwstra, Dijck-Brouwer and Wildeman6). Mothers of 314 infants chose to bottle-feed their child and 160 opted for breast-feeding. The infants receiving formula were randomised into a standard formula group (control formula, CF; n 169) and a LCPUFA-supplemented formula group (LF; n 145). Standard formula consisted of Nutrilon Premium® (Nutricia, Zoetermeer, The Netherlands). For the supplemented formula, the lipid fraction of Nutrilon Premium® was enriched with 0·45 % (by weight) arachidonic acid and 0·30 % (by weight) DHA. The duration of supplementation was 2 months. In case breast-feeding stopped before 2 months, the infant received LCPUFA-supplemented formula until the age of 2 months, and remained in the BF group. All formula-fed infants received control formula between 2 and 6 months. The children underwent neurodevelopmental assessment at 3 and 18 months of age(Reference Bouwstra, Dijck-Brouwer and Wildeman6, Reference Bouwstra, Dijck-Brouwer and Boehm7). Follow-up was achieved in 84 % (3 months(Reference Bouwstra, Dijck-Brouwer and Wildeman6)) and 92 % (18 months(Reference Bouwstra, Dijck-Brouwer and Boehm7)) of the original groups. All children seen at the 18-months follow-up were eligible for re-testing at 9 years. At the 9-year follow-up, both parents and examiners were unaware of the type of formula-feeding the infant had received. The examiners also were blind to the type of milk fed during the first 8 weeks.
Procedures
Neurological condition of the children was evaluated with the Touwen (1979) examination(Reference Touwen8), which is a standardised, age-specific assessment designed for the assessment of MND. Essential in the diagnostics of MND is the presence of coherent clusters of signs. Single signs do not have clinical significance; signs only have significance when they co-occur (cluster) with other signs within a functional domain. The examination is organised into eight functional domains: posture and muscle tone; reflexes; dyskinesia; coordination; fine manipulative ability; associated movements; sensory deficits; cranial nerve functioning. The examination results in a clinical classification: normal; simple MND; complex MND; or abnormal. A child is considered neurologically abnormal in the presence of a clear neurological disorder such as cerebral palsy. Simple MND denotes the presence of one or two domains of dysfunction and is present in about 15–20 % of children. It has little clinical relevance and can be regarded as typical but non-optimal brain functioning, in other words as a minor neurological difference. Complex MND denotes the presence of more than two domains of dysfunction and is the clinically relevant form of MND. In an aetiological sense it can be considered a borderline form of cerebral palsy as it is linked to a chain of pre- and perinatal adversities(Reference Hadders-Algra3, Reference Peters, Maathuis and Kouw9). A child is classified as neurologically normal when no domains are scored as deviant or in case of the isolated presence of a mild dysfunction in reflex activity.
The neurological examination according to Touwen(Reference Touwen8) has a good intra-rater, inter-rater and test–retest reliability; the κ statistics of the three forms of reliability for neurological classification ranged between 0·71 and 0·83(Reference Peters, Maathuis and Kouw9). Its construct validity is reflected by the differential relationship for simple and complex MND with prenatal and perinatal adversities: adverse conditions during early life have a weak to moderate relationship with simple MND and a strong correlation with complex MND(Reference Hadders-Algra3, Reference Hadders-Algra, Huisjes and Touwen10). Predictive validity is good; this is reflected by the relationship between the severity of MND at 9 years and the risk of MND at 12 and 14 years and that of learning and behavioural problems at 9 and 14 years(Reference Hadders-Algra3, Reference Soorani-Lunsing, Hadders-Algra and Olinga11, Reference Soorani-Lunsing, Hadders-Algra and Huisjes12).
The study's primary outcome parameter was the NOS. The NOS uses the optimality concept to summarise neurological condition and provides a sensitive measure of the child's overall neurological status. The sensitivity of the NOS to detect effects of early nutrition may be illustrated by the study of Bouwstra et al. (Reference Bouwstra, Dijck-Brouwer and Decsi13) which demonstrated that prenatal fatty acid status was related to NOS at 18 months, but not to outcome measured with the Bayley Scales of Infant Development. For sixty-four items, representing the entire neurological examination, an optimal range was identified (see online Supplemental Table 1). The total number of items with a value within the predefined optimal range formed the NOS of a child. It should be realised that there is a conceptual difference between normality and optimality, as the range for optimal behaviour is narrower than that of normal behaviour(Reference Prechtl14). The NOS may be regarded as a quantitative and more subtle expression of the clinical neurological condition. This also means that the two measures are highly correlated(Reference Touwen, Huisjes and Jurgens-Van der Zee15) (present study ρ = − 0·691; P < 0·001). The quantitative and precise nature of the NOS makes it a suitable instrument to evaluate subtle deviations in neurodevelopmental outcome. The NOS has previously been used in infancy and pre-school-age children; in the present study NOS principles are applied for the first time at the age of 9 years.
Data on prenatal and perinatal conditions had been collected during enrolment with the help of the Obstetrical Optimality Score (OOS). The OOS describes the obstetrical conditions ranging from the parents' socio-economic status to the infant's condition immediately after birth(Reference Touwen, Huisjes and Jurgens-Van der Zee15). At the 18-months assessment, maternal verbal intelligence quotient (IQ) was estimated using a very abbreviated version of the Wechsler Adult Intelligence Scale, WAIS III, limited to the subtests information and vocabulary(Reference Wechsler16). Social condition was documented with the Home Observation for Measurement of the Environment (HOME) inventory(Reference Caldwell and Bradley17). The HOME contains forty-five items clustered into six subscales: Parental Responsivity; Acceptance of Child; Organization of the Environment; Learning Materials; Parental Involvement; Variety in Experience. At the 9-year follow-up information was collected on parental education and profession, the child's medical history, family composition and nutritional habits.
Depending on the wish of the participants, the assessment was carried out in the hospital or at home. The present study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Ethics Committee of the Groningen University Hospital. Written informed consent was obtained from all subjects. The trial is registered under ISRCTN52788665.
Statistical analysis
Statistical analysis focused on differences in neurological outcome between the two randomised formula groups. In addition, differences between the breast-fed group and the formula groups were analysed. Specific attention was paid to the effect of site of investigation and maternal estimated verbal IQ. Neurological classification, specific clusters of dysfunction and NOS were not normally distributed. Univariate analyses were performed with the Mann–Whitney U test and the Kruskall–Wallis test.
In order to analyse the effect of nutritional group on neurological classification and type of dysfunction while taking into account the role of potential confounders, logistic regression analyses were performed. The following confounders, associated with neurological outcome at P < 0·20, were entered into the analyses: sex; maternal smoking during pregnancy; duration of the second stage of delivery; birth weight; OOS; maternal hypertension during pregnancy; Apgar score 3 min after birth; maternal educational level. To investigate the effect of nutritional group and the above-mentioned confounders on NOS, multiple linear regression was applied to the NOS transformed to the fifth power. This transformation to normality was identified by the Box–Cox method applied to regression model residuals (the NOS was skewed to the left).
The multivariate analyses on the effect of nutritional group were carried out in two steps. In the first analysis the effect of the type of formula was assessed; in the second analysis the difference between formula- and breast-feeding was evaluated. P ≤ 0·05 was considered statistically significant. Statistical analyses were performed using SPSS 14.0 for Windows (SPSS, Inc., Chicago, IL, USA).
Results
All children who were tested at 18 months were invited for follow-up at 9 years (n 436); 341 of the 474 children of the original study agreed to participate, ninety-one children in the LF group (63 %), 123 children in the CF group (73 %) and 127 children in the BF group (79 %; Fig. 1). Obstetrical data of the study groups and sociodemographic characteristics of the parents are described in Table 1. In general, obstetrical and social characteristics of the children who were and who were not assessed at 9 years were comparable. Yet, children who were not assessed at 9 years had shown significantly more often normal-optimal general movements at 3 months than those who had been assessed, meaning that children with an optimal neuromotor condition in early infancy were under-represented in the follow-up of the three groups at 9 years (P = 0·003). In addition, attrition in the LF group was more selective than in the other groups. First, in the LF group more boys did not take part in the follow-up (thirty-five boys and seventeen girls) compared with the BF group (fifteen boys and seventeen girls) and the CF group (twenty-four boys and seventeen girls). Second, the mental developmental index of the Bayley Scales of Infant Development(Reference Bayley18, Reference van der Meulen, Ruiter and Lutje Spelberg19) at 18 months of the children of the LF group who were not assessed at 9 years was significantly lower than that of LF children who did participate in the follow-up at 9 years (P = 0·007). A similar selective attrition was not present in the CF and BF groups.
Flow diagram of children enrolled in the study and followed up until 9 years of age. * for more detailed information, see Bouwstra et al. (Reference Bouwstra, Dijck-Brouwer and Boehm7). LCPUFA, long-chain PUFA.
Obstetrical and social characteristics of the three groups assessed at 9 years
BF, breast-fed; LF, long-chain PUFA-supplemented formula-fed; CF, control formula-fed; HOME, Home Observation for the Measurement of the Environment inventory; OOS, Obstetrical Optimality Score.
* Significant difference between all groups (P = 0·001).
† Significant difference between BF and LF/CF combined (P = 0·003).
Neurological condition at 9 years was not affected by the site of investigation (in the hospital or at home), by current consumption of fish (dichotomised as at least once per week or less than once per week) and maternal verbal IQ. The NOS of children in the LF and CF groups did not differ; in both groups the median NOS score was 57. The NOS of formula-fed children (LF and CF groups) was significantly lower than that of breast-fed children (median values 57 and 58, respectively, Mann–Whitney, P = 0·008; see also Table 2 for more specifics on the BF group). Multiple regression (Table 3) confirmed that the NOS was not affected by LCPUFA supplementation. The analysis revealed an interaction between sex and type of feeding: girls who had been breast-fed had a higher NOS than girls who had been formula-fed. Other factors associated with a higher NOS were higher birth weight and an OOS above the 10th percentile. The effects presented in Table 3 are on the transformed scale of NOS. An effect of size, for example, 0·4 on the transformed scale can be interpreted on the original scale as follows. A NOS value of 57 transforms to (57/50)5 = 1·9254; increasing this by 0·4 gives 2·3254; transforming this back to the original scale gives 50 × 2·32 540·2 = 59·2. Note that as the width of most confidence intervals in Table 3 is below 0·4, the CI width thus becomes about 2 points on the original NOS scale. The NOS at 9 years showed a statistically significant association with the mental developmental index of the Bayley Scales of Infant Development at 18 months (Spearman ρ = 0·193; P < 0·0001).
Duration of exclusive breast-feeding and Neurological Optimality Score
Results of linear regression analysis of factors contributing to the normalised Neurological Optimality Score (NOS): (NOS/50)5*
LF, long-chain PUFA-supplemented formula-fed group; CF, control formula-fed group; BF, breast-fed group; OOS, Obstetrical Optimality Score.
* n 330; R 2 0·14; intercept for a CF male with birth weight 3·5 kg and OOS above 10th percentile, 1·7265.
Of the children, 46 to 54 % had a normal neurological condition, 36–39 % showed simple MND and 10–15 % of children had complex MND. Neurological classification did not differ between the randomised formula groups. Children in the BF group tended to have a slightly better neurological condition than the children in the formula-fed groups, but the difference did not reach statistical significance (P = 0·17; Table 4).
(Number of children and percentages)
BF, breast-fed; LF, long-chain PUFA-supplemented formula-fed; CF, control formula-fed; MND, minor neurological dysfunction.
* No statistically significant differences were present between the groups.
LF and CF groups did not differ in the prevalence of specific types of dysfunction. Children of the BF group, however, showed significantly less often fine manipulative dysfunction than formula-fed children (BF 16 %, LF 30 %, CF31 %; BF v. formula groups; P = 0·002). This was confirmed in multivariate analysis (Table 5) with the following contributing factors: OOS below the 10th percentile; male sex; maternal smoking during pregnancy; duration of the second stage of delivery. Breast-fed and formula-fed children did not differ in the prevalence of other types of neurological dysfunction.
(Odds ratios and 95 % confidence intervals)
BF, breast-feeding; LF, long-chain PUFA-supplemented formula-feeding; CF, control formula-feeding; OOS, Obstetrical Optimality Score.
Discussion
The present study indicated that consumption by term infants of formula containing LCPUFA for the first 2 months after birth did not affect neurological condition at 9 years. The study also revealed that fine manipulative ability of breast-fed children was better than that of formula-fed children.
A major limitation of the study is its attrition. Overall attrition was 28 %, which – over a period of 9 years – can be regarded as relatively favourable(Reference Fewtrell, Kennedy and Singhal20). However, a major problem in the present study was the selective nature of attrition, that is, a selective loss of boys and children with a worse cognitive development at 18 months in the LF group. The selective attrition interfered with the randomised design of the study. While the multiple regression analysis can alleviate the effect of attrition if the missingness of data depends only on known covariates, it cannot annihilate the effect of severe (‘missing not at random’) attrition.
Some of the breast-fed children received a few weeks of formula supplemented with LCPUFA (Table 2). This may be considered a limitation, as it may have influenced any differences between the supplemented formula group and the breast-fed group. However, the finding that the neurological condition of the LF group did not differ from the CF group reduces the likelihood that differences between the LF and BF groups were affected by the mixed composition of the BF group.
The prevalence of MND in the present study is another point which deserves methodological attention. The prevalence is higher than indicated by earlier estimations of MND in the general population(Reference Hadders-Algra3). This may raise questions regarding the representativeness of the current sample of healthy full-term infants. However, neurological evaluation of these groups at 3 months indicated that neurological status was representative for the general population(Reference Bouwstra, Dijck-Brouwer and Wildeman6, Reference Bouwstra, Dijk-Stigter and Grooten21). Possibly, the relatively high prevalence of MND in the present study reflects the general trend of worsening neuromotor condition in the last decades(Reference Hadders-Algra22).
The strengths of the study are its randomised design and its assessor-blinded evaluation with an internationally recognised, sensitive technique to evaluate neurological condition(Reference Lanting, Fidler and Huisman4, Reference Barnett23–Reference Arnaud, Daubisse-Marliac and White-Koning26). Based on the width of the confidence intervals, it is concluded that the LF v. CF differences do not exceed 2·2 points on the NOS scale.
This is the first study reporting the effect of supplementation of formula with LCPUFA in healthy term infants on neurological condition at school age. Using the NOS, a sensitive indicator of neurological condition, an effect of 2 months of postnatal LCPUFA supplementation on neurological status at 9 years could not be demonstrated. Bearing in mind the fact that the LF group suffered from selective attrition of children with a lower mental developmental index at 18 months, the present study indicates that LCPUFA supplementation during the first 2 postnatal months does not promote neurological condition at school age.
The present finding of a subtle positive association between breast-feeding and neurodevelopmental outcome, and neurological condition in particular, is in line with reports of others(Reference Lanting, Fidler and Huisman4, Reference Anderson, Johnstone and Remley27, Reference Lanting, Patandin and Weisglas-Kuperus28). The association between breast-feeding and neurological outcome was less strong than previously reported by Lanting et al. (Reference Lanting, Fidler and Huisman4), which may be attributed to differences in the populations studied (Lanting et al. study: a mix of high-risk and low-risk infants; present study: healthy full-term infants) and the quality of the breast-feeding data (Lanting et al. study: retrospectively collected information; present study: detailed prospective information). Der et al. (Reference Der, Batty and Deary29) were able to demonstrate that maternal IQ accounted for a major part of the association between breast-feeding and developmental outcome in terms of IQ. In the present study, a subtle association between breast-feeding and fine manipulative ability remained, also when estimated maternal verbal IQ was taken into account. Fine manipulative ability is mediated especially by cortical–subcortical networks. These networks do not only play a role in sensorimotor aspects of motor programming, movement planning, program selection and motor memory but also in cognitive functions, such as intelligence(Reference Alexander and Crutcher30, Reference Diamond31). This means that the association between breast-feeding and better fine manipulative ability corresponds to reports of others of an association between breast-feeding and higher IQ(Reference Anderson, Johnstone and Remley27). Interestingly, the NOS data indicated that especially girls profit from the beneficial effect of breast-feeding. This sex-specific effect may be explained by shared genome, and, albeit less likely, sex-specific differences in metabolism of α-linolenic acid: adult women have a higher conversion rate of α-linolenic acid to EPA and DHA than adult men(Reference Burdge32).
In conclusion, the present study indicates that LCPUFA supplementation of formula during the first 2 postnatal months in healthy term infants does not promote neurological condition at school age. In addition, the study confirmed that breast-fed infants have a slightly better neurodevelopmental outcome than formula-fed infants – reflected in the present study by a reduced prevalence of fine manipulative dysfunction. Finally, the study underscores the need of the evaluation of selective attrition with respect to early developmental data in studies assessing the effect of early nutrition on long-term developmental outcome.
Acknowledgements
We gratefully acknowledge the critical comments of Dr G. Boehm on a previous draft of the paper.
Previous parts of the LCPUFA study have been funded by Numico Research B.V. The follow-up at 9 years is part of the Early Nutrition Programming Project (EARNEST), which is funded under the Food Quality and Safety Priority of the Sixth Framework Programme for Research and Technical Development of the European Community (FOOD-CT-2005-007036; www.metabolic-programming.org).
M. H.-A. initiated the follow-up study; C. de J. and H. K. K. collected the data; C. de J., V. F. and M. H.-A. analysed the data; C. de J. and M. H.-A. drafted the report; H. K. K. and V. F. commented on drafts. All authors have seen and approved the final version.
There are no conflicts of interest.
Supplemental Table 1 is available online only at http://journals.cambridge.org/action/displayJournal?jid=bjn
