To investigate the effects of dietary supplementation with 600 mg/day of the omega-3 LC-PUFA docosahexaenoic acid (DHA) on children's reading, working memory and behavior over a 16-week period.

The hypothesis was that DHA would be of benefit in each of these domains.

The study was open to healthy children attending any mainstream Oxfordshire primary school who were in the third, fourth or fifth year-groups. Such children are typically aged 7–9 years, although a minority is aged 6 or 10 years.

As originally designed, the inclusion criteria required children to be below the 20th centile on a age-standardized word reading test normed on UK children [15]. In children within this age range, this would typically equate to a reading performance of around 2 years below the level expected for the child's chronological age. Before first randomization, however, it became apparent that planned study numbers would not be achieved unless this inclusion criterion was relaxed to the 33^rd centile (equivalent to reading at around 18 months behind chronological age); hence, the protocol was modified accordingly.

Children with specific medical disorders (e.g. visual or hearing impairment), general learning difficulties, or who were taking medications expected to affect behavior and learning, were excluded from the study, as were those whose first language at home was not English. Schools were also asked to exclude any children whose social/family circumstances would have made inclusion into the study inappropriate (e.g. serious illness in the family). Children who, according to their parents, ate fish more than twice a week or took omega-3 supplements were also excluded.

The Oxfordshire Local Authority was an active partner in the research, and they provided information on how children from participating schools had performed on the national attainment tests conducted on all 7 year olds in state schools in England, Wales and Northern Ireland [16]. From this, an initial list was drawn up of all children with below average attainment in reading at that age. Teachers at participating schools were then invited to modify this according to their opinions of the children's current reading performance. On this basis, letters were sent to parents inviting their children to take part in the formal screening assessments, in which their reading ability was individually tested to determine eligibility (see Figure 1).

Written informed consent was gained from parents, and verbal assent from the children, prior to the initial screening assessments. The protocol for this trial and a CONSORT checklist are included as supporting information. Ethical consent was gained from the Milton Keynes Research Ethics Committee (08/H0603/49).

Children in the Active group received a fixed dose of 600 mg DHA (from algal oil), delivered in three 500 mg capsules per day, each providing 200 mg DHA. The Placebo treatment consisted of three 500 mg capsules per day containing corn/soybean oil, matched with the Active treatment for taste and color. Both treatments were provided by Martek Biosciences Corporation (now DSM Nutritional Lipids).

Schools were given a 16 week supply of capsules (labelled with each participating child's name) and asked to dispense 3 capsules to all participating children once a day at lunch time during school terms. Parents were also given a 16-week supply of capsules and asked to dispense these to their children at weekends, during school holidays and at any other time when their children were not in school. Both schools and parents were given full instructions for dispensing capsules and a diary to record capsule consumption, aimed at both monitoring and encouraging children's compliance. Both schools and parents were encouraged to telephone the research team at any time if they had any concerns.

Primary outcomes, assessed at the baseline screening appointment for all children, and again at 16-week follow-up for the children randomised, were as follows:

Power calculations were performed based on the only previous RCT to assess both reading ability and behavior in children [7]. The results indicated that 180 participants per group would provide 90% power with an α of 5%, for an effect size of r = −0.169 (Cohen's d = −0.343).

A statistician at the Centre for Statistics in Medicine in Oxford independently performed the randomization. The program included a minimization algorithm to ensure balanced allocation of participants across the treatment groups for school (to control for possible between-school differences) and sex (a potentially important factor) [20]. Randomisation was performed only after eligibility was assured, and it was concealed until after the initial two-group analyses were complete.

Investigators, participants and those assessing outcomes were all blind to treatment allocation. Post-intervention, both teachers and parents of participants were asked whether they thought their child had been allocated to Active treatment or Placebo, and these estimates were used to assess the maintenance of blinding.

Group comparisons were carried out on primary outcomes using change scores (i.e. the post-intervention score minus baseline score), in line with previous studies. These were considered likely to be more sensitive to treatment effects than total scores, given the short intervention period of 16 weeks. Main analyses were conducted on an intention-to-treat principle (ITT): thus, all children were included according to treatment allocation, irrespective of compliance with the intervention. Appropriate checks were made that participants with missing data did not differ significantly on any demographic variables. Missing data were imputed using median values by treatment group. For all primary outcomes, additional planned group comparisons were carried out on the subgroup of children whose baseline reading scores were ≤20^th centile (based on the original protocol), and also on those ≤10^th centile (to evaluate any possible trends related to the severity of initial reading problems).

As recommended for the credibility of subgroup analyses [21], tests of interaction were also carried out to assess potential differences in treatment efficacy between the two main subgroups, i.e. all children with initial reading ≤20^th centile and those children with milder degrees of initial reading impairment.

For any measures with more than 15% missing data, additional per-protocol analyses were conducted i.e. including only those participants with complete data.

Recruitment was carried out in 74 Oxfordshire mainstream primary schools beginning in January 2009 and finishing in November 2010. Post-intervention assessments (16 weeks after enrolment) were completed in April 2011. Of the 1376 children who were invited, 675 of their parents/guardians gave consent and their children were assessed. Of these, 362 met study inclusion criteria and were randomized. The most common reasons for exclusion were that their reading exceeded the 33^rd centile (n = 236), followed by eating fish more than twice a week (n = 50); other reasons for exclusion are described in the flowchart of participants (n = 27) detailed in Figure 1.

Of the 362 children randomized, 359 were assessed again after the 16-week intervention (179 Active, 180 Placebo). Three children were lost at follow up (1 in the Active group, 2 in the Placebo group). In addition, one child from the Placebo group failed to complete the Recall of Digits assessment at the post-intervention follow-up.

The two treatment groups did not differ on any of the core demographic variables, nor on any of the primary outcome measures at baseline. Demographic information is provided in Table 1. The mean age of the sample was 8 years 8 months, 53% were male, 91% were white, and just over 20% were eligible for free school meals (our proxy for low socio-economic status). Baseline data on the primary outcomes are shown in Table 2. With respect to these, mean reading performance of the children randomized was 1.5 sd below normal, equating to a reading performance around 18 months below chronological age. Working memory scores were around 0.5 to 1 sd below population norms as derived from the British Ability Scales II [15]. On the behavior measures (where higher scores indicate greater difficulties), both teacher and parent ratings were all within the normal range, with the exception of the ‘cognitive problems’ sub-scale (assessing attentional and related difficulties), where these children scored 1 sd above population means. A few other scales showed slight elevations (>+0.5 sd), including ‘oppositional’ (both parents and teachers), ‘anxious-shy’ (teacher ratings only) and both ‘social problems’ and ‘psychosomatic’ (parent ratings only).

Parent and teacher estimates of group allocation at post-intervention were used to assess the maintenance of blinding. Group comparisons carried out on these estimates showed there were no significant differences between groups, as shown in Table 3.

Intention-to-treat analyses were carried out on all children randomized (n = 362) and on the pre-planned sub-groups defined by baseline reading ≤20th and 10th centiles (n = 224 and n = 105 respectively) in line with the original protocol. Behavior ratings were the only measures with >15% of the data missing (change scores n = 273 for Teachers, and n = 246 for Parents), so additional per protocol analyses were conducted on these measures.

The ITT analyses showed that behavioral improvements as rated by parents were greater for Active treatment over Placebo on six of the seven CPRS-L Global scales (Table 7). Their means (sds) were as follows: ADHD Index (Active −4.5 (8.9), Placebo −2.6 (7.4), p<0.04), Restless-Impulsive (Active −4.8 (8.6), Placebo −2.3 (8.2), p<0.01), Emotional Lability (Active −3.1 (9.5), Placebo −0.2 (9.3), p<0.01), Global Index Total (Active −4.3 (8.8), Placebo −1.8 (8.2), p<0.01), DSM-IV Hyperactive-Impulsive (Active −3.8 (8.3), Placebo −2.2 (8.2), p<0.02), and DSM IV Total (Active −3.8 (8.1), Placebo −2.1 (7.4), p<0.03). Significant differences in favor of Active treatment were also found for two of the seven CPRS-L sub-scales: Oppositional (Active −3.2 (10.1), Placebo −0.4 (8.1), p<0.01), and Hyperactivity (Active −2.9 (7.2), Placebo −1.2 (7.1), p<0.01).

In the per protocol analyses (n = 247), a similar pattern of results was found (Table 8), although differences reached statistical significance only for Anxiety (Active −4.1 (8.5), Placebo −2.0 (7.7, p = <0.05), Global Restless-Impulsive (Active −4.1 (7.5), Placebo −2.2 (6.8), p<0.02), Global Emotional Lability (Active −2.4 (7.7), Placebo −1.0 (8.3), p<0.03) and Global Index Total (Active −3.8 (7.1), Placebo −2.0 (6.8) p<0.02).

No significant effects of treatment were found within the sub-groups defined by severity of reading impairment at baseline. (These data are not shown, but can be provided on request).

The ITT analyses showed no significant differences in behavior change scores between Active and Placebo groups on any of the CTRS-L sub-scales or global scales, with the exception of the ‘Perfectionism’ sub-scale, on which both groups scored below the normal population mean at baseline and post-intervention. (Active mean = 0.2, sd = 6.5, Placebo mean = −0.8, sd = 7.4, p<0.03). In the per protocol analyses (n = 273), no group differences were significant.

No significant effects of treatment on teacher-rated behavior change were found in the sub-groups defined by severity of reading impairment at baseline. (These data are not shown, but can be provided on request).

The DHA supplement used is generally regarded as safe (GRAS) and so no stopping guidelines were put in place except in the case of serious adverse events. As expected, there were none in the course of this trial. The parents of three children reported minor adverse events during the intervention period and so they discontinued treatment, but no unblinding was required. The adverse events were: asthma symptoms (Active), nettle rash (Placebo) and disruptive behavior (Active). The parent of one child in the Active group reported hair loss 6 weeks after completing the study.

Counts of capsules returned by schools indicated mean compliance of approximately 75% and this did not differ between Active (n = 138) and Placebo groups (n = 157). (From 200 capsules allocated to schools for each child, quantities returned were: Active mean = 50.0, sd = 51.8) and Placebo Mean = 42.0, sd = 39.5). Of the 50 capsules allocated to parents for non-school days, more than 50% of data were missing and so these are not reported.

The context in which this trial was carried out was mainstream UK schools, where the influence of the researchers was low and contact with parents was indirect and minimal, whereas most previous trials have involved children with specific behavioral and/or learning difficulties recruited either from clinics or from direct advertisement to parents. This classifies it as an “effectiveness” - i.e. ‘real-world’ study - rather than an “efficacy” study, and provides reasonable grounds to trust the generalizability of these data to other mainstream school populations, albeit with some caveats.

Generalizability to all children aged 7–9 years attending mainstream UK schools is clearly limited, as our sample was pre-selected for underperformance in reading. In terms of socioeconomic status, the children in this study were fairly representative of the general population of England, as the percentage receiving free school meals was comparable to national figures (20.2% vs 18.6%). However, ethnic minorities were under-represented (91.2% were white, compared with 77.7% nationally), primarily because our study required that children use English as their first language at home. The percentage of boys was also slightly elevated (53% vs 51% nationally), reflecting the fact that in this age range, boys are typically more likely than girls to be underperforming in reading.

Various implementation issues concerning dose and formulation, delivery and uptake may have a bearing on these results and therefore merit brief consideration.

The dose we selected (600 mg/day) was comparable to those used in most other trials of long-chain omega-3 for child behavior and/or learning to date [6], and only slightly higher than the intake of 500 mg/day recommended for general cardiovascular health in adults by leading scientists in the field [29]. Higher doses might possibly yield more substantial treatment effects, but this would need to be investigated via systematic dose-ranging studies, because intakes of individual nutrients may not relate in any simple linear fashion to health or performance outcomes [30].

The formulation we used provided only DHA, whereas most previous studies have used varying combinations of DHA, EPA and sometimes other fatty acids. A recent meta-analysis found that treatment effects on ADHD-type symptoms increased with the proportion of EPA (and thus decreased with the proportion of DHA) in the supplements used [6]. This analysis was based on only 10 small and highly heterogeneous trials, however, and direct comparisons would be needed to evaluate the relative merits of different formulations with any degree of certainty.

The efficiency of delivery and uptake can also influence results in any intervention study. The supplements were provided to children on school days by a range of different staff at the 74 schools involved. Counts of capsules returned from schools suggested that satisfactory compliance with the treatment was achieved in these settings, but these data were not complete. Similarly, the parents who gave the children their supplies during weekends and other non-school days will have varied in their motivation and adherence to the treatment schedule; compliance data here were unfortunately impossible to obtain, in line with other studies.

This study provides the first evidence that dietary supplementation with the omega-3 DHA might improve both the behavior and the learning of healthy children from the general school population. The supplement was found to be safe and well tolerated, as expected. The benefits we found for reading were only evident amongst those children whose initial reading performance fell within the lowest 20% of the normal distribution, suggesting that for this particular outcome measure, DHA supplementation should be regarded as a targeted intervention for the poorest readers rather than a universal one. For behavior, however, it is possible that the benefits may have broader applicability, as reductions in parents' ratings of ADHD-type symptoms were evident across the entire sample studied here, despite the relatively mild nature of these children's initial behavior problems. The current results thus extend the previous findings in this area from clinically-defined groups to healthy but underperforming children from the general school population, and suggest that dietary deficiencies of DHA might have subtle behavioral effects on children in general, as has been shown for some combinations of artificial food additives [31].

Dietary intakes of long-chain omega-3 in modern western-type diets are widely acknowledged to be sub-optimal, both for general physical health and for mental health and performance [32]. Similarly, the importance of early intervention for behavior and literacy problems in children is widely acknowledged to be more effective and less costly than interventions later in life, by which time such problems can become compounded and the life chances of these children further compromised [11]. These findings clearly require replication, but we suggest that future studies should adopt a similar focus on the most vulnerable groups within the general population, ideally using additional measures to assess the relative costs and benefits of this kind of dietary supplementation.