Background
The World Health Organization (WHO) defines neurodevelopmental disorders as one of today’s greatest public health challenges [
1], with Attention Deficit Hyperactivity Disorder (ADHD) being one of the most prevalent among children in Norway and worldwide [
2,
3]. ADHD is characterized by a number of behavioural traits, including inattention, impulsivity and hyperactivity [
4] and it typically co-occurs with delayed language development, motor functions, impaired emotional control as well as with other psychiatric disorders [
5,
6]. Furthermore, impairments related to ADHD often persist into adulthood [
7‐
10], alongside poorer social functioning and an increased risk of unemployment [
3,
6].
ADHD aetiology is multifactorial and complex [
6], involving a genetic contribution [
6,
11], as well as social- and environmental factors [
12]. Because treatment options available for young children are limited [
13,
14] and due to the relatively poor long-term effectiveness of medical treatment for ADHD [
15], identification of early-life modifiable risk factors, such as nutritional factors, could be a key strategy in improving prevention approaches for ADHD in children [
13,
16].
Findings from investigations into associations between child diet and ADHD have been equivocal. Some research on ADHD patients have found suboptimal levels of several nutrients, including zinc [
17], iron [
18], magnesium [
19] and Omega-3 [
18,
20], but dietary interventions aiming to reduce symptoms have yielded mixed results [
21,
22]. A recent meta-analysis reviewed the scarce literature investigating dietary patterns and ADHD diagnosis or symptoms in children [
23]. Overall, the authors found higher risk of ADHD diagnosis or symptoms for children having low scores on a healthy dietary pattern and high scores on an unhealthy dietary pattern. However, all studies were observational, where half investigated associations cross-sectionally, which severely impedes a causal interpretation of these associations.
It has been suggested that an optimal maternal diet during pregnancy might be even more crucial to child ADHD outcomes than early childhood diet [
15]. Surprisingly, however, the possible impact of maternal diet quality during pregnancy on child neurodevelopmental functions related to ADHD has been largely neglected until recently. The increased interest in this area is likely due to the evidence that the in-utero environment might be vital in the development of neurodevelopmental disorders [
24‐
26]. Research on the effects of severe and prolonged maternal malnourishment on brain development in offspring highlights the important role of maternal diet [
18], particularly during the prenatal and early postnatal period [
6]. It remains less clear how subtle differences in maternal diet quality during pregnancy might influence the trajectory of child neurodevelopmental disorders.
We have not identified any research investigating overall maternal diet quality during pregnancy in relation to child ADHD diagnosis. Moreover, the limited amount of literature investigating prenatal and child diet in relation to neurodevelopmental problems associated with ADHD has indicated that there exists a possible association [
27,
28]. However, these associations tend to be either small or from smaller studies [
29]. Hence, there is a need for large population-based studies that investigate longitudinal associations between prenatal and early postnatal diet quality and ADHD.
The aim of this paper is to investigate the associations between both maternal diet quality during pregnancy and child diet quality, using diet quality indices, and maternally reported child ADHD symptoms at 8 years and child ADHD diagnosis retrieved from the Norwegian Patient Registry.
Discussion
In this study, we investigated the associations between maternal and child diet quality and maternally reported ADHD symptom score at 8 years and child ADHD diagnosis. We observed inverse associations between maternal diet quality during pregnancy (measured by the PDQI) and both child ADHD symptom scores and child ADHD diagnosis, and a positive association between the contribution of ultra-processed food to total maternal energy intake in pregnancy (measured by the UPFI) and child ADHD symptom scores. We found no reliable association between UPFI and child ADHD diagnosis, or between child diet quality (measured by the CDQI at 3 years) and either outcome.
The difference in results for maternal diet quality during pregnancy and child diet quality at 3 years might be due to difference in measurement detail, since the maternal FFQ is more extensive than the questions about food intake in the child questionnaire (255 vs. 36 food items), which is likely to influence the specificity and accuracy of the instrument and subsequent dietary index scores calculated from those instruments. Nevertheless, we acknowledge that the child dietary questions include the main foods and beverages consumed at three years of age.
If the associations between maternal diet quality during pregnancy and child diet quality and ADHD symptoms and diagnosis are due to causal effects from maternal diet quality on child development, these results would support the idea of prenatal programming. During critical stages of the prenatal period, the foetus may be particularly sensitive to environmental influences like maternal nutrition, and the subsequent insults based on these influences may sustain long after birth and into adulthood [
26]. A proposed mechanism behind prenatal programming is epigenetic modifications [
25,
63‐
65] via e.g. immune activation [
24,
66,
67]. Interestingly, one study has found an association between prenatal high fat/high sugar diet and ADHD in youth with early onset conduct problems indirectly via epigenetic modifications of the insulin-like growth factor gene [
68]. Still, studies that use robust causal identification strategies to ensure that observed associations between diet and ADHD are due to causal effects of diet are an important next step.
We found no associations between child diet quality at 3 years and either outcome, which adds to the so far undecided evidence base investigating child diet and ADHD. A recent meta-analysis found healthy dietary patterns to be associated with low ADHD symptoms in children (3–11 years old) and adolescents (12–16 years old) and unhealthy dietary patterns a risk factor for ADHD, but the authors emphasize that due to limitations in study design across most included studies, these associations do not constitute evidence for causal effects [
23]. Also, the authors do not explain how they dealt with dependencies in the effect estimates, nor if they investigated for sources of heterogeneity other than publication bias, which indicates that further caution should be made in making firm inferences from their findings. Conversely, Mian et al. [
69] found ADHD symptoms at 6 years to be predictive of child diet quality assessed at 8 years, but not vice versa, with ADHD symptoms at 10 years, suggesting that postnatal diet quality does not influence ADHD symptoms, but rather that some ADHD symptoms, e.g. low levels of impulse control, may result in a poorer diet.
We found associations of similar magnitude for both maternal diet quality indices in relation to child ADHD symptom score, however the associations were small. This is in line with the small association estimates found in a previous review and meta-analysis investigating maternal diet quality and child neurodevelopmental outcomes. More recently, Mesirow and colleagues [
70] studied aspects of maternal diet quality in relation to child behavioural issues in children with either low conduct problems or early-onset persistent conduct problems. They found that, specifically for children presenting with early-onset conduct problems, mothers had a poorer diet quality (lower fish intake, higher processed food consumption) compared to mothers with children having low conduct problems. Additionally, maternal processed food consumption was associated with higher childhood hyperactivity (4–10 years) independent of conduct problem trajectories, but the effect estimate was small.
Isaac and Oates [
71] propose that large effect estimates cannot be expected in generally healthy populations with adequate diet quality, such as the MoBa sample which consists of, on average, well educated mothers in a rich, industrialized country, when investigating outcomes related to child developmental functions. Isaac and Oates further argue that even though results might indicate an association, there might be no clinical significance as the absolute effect estimates might be minute [
71]. This increases our confidence in proposing that associations seen between the maternal diet quality indices and child ADHD symptoms in generally healthy populations with, on average, adequate diet quality are so small that they have no clinical relevance. However, small effects have the potential to influence outcomes at the population level, particularly in the situation where the exposure, in this case diet, affects the entire population. Still, more research is needed to understand the association between maternal diet and child ADHD symptoms and diagnoses in populations with a higher prevalence of inadequate maternal diet quality during pregnancy.
For child ADHD diagnosis, we only found associations with the PDQI (indicates adherence to the Norwegian food-based dietary guidelines) and not for the UPFI (indicates consumption of ultra-processed foods). It is likely that the ADHD diagnosis outcome (retrieved from patient registry) is less afflicted by bias than the ADHD symptom score outcome (based on parent reported symptoms), which might be one reason for the difference in results for the two outcomes. If we assume that the results for ADHD diagnosis are more accurate due to less presence of bias, compared to the associations seen for the ADHD symptom score outcome, it can be an indication of overall maternal high diet quality during pregnancy being more important for ADHD development than the proportion of intake of ultra-processed foods, which are mostly of low nutritional value. This is in line with recent findings investigating maternal dietary patterns and birth outcomes, which found stronger associations with preterm birth for the healthy pattern compared to the unhealthy, and the opposite trend for birth weight [
72], but only for data driven patterns, and not a-priori defined indices that are more generalizable. Looking at studies with outcomes that are related to ADHD, Jacka et al. [
73] found the opposite trend, wherein the unhealthy dietary pattern showed a stronger association with symptoms of child behavioural difficulties compared to the healthy dietary pattern. Conversely, Steenweg-de Graaff et al. [
74] found similar strengths of associations for low adherence to a healthy dietary pattern and high adherence to an unhealthy dietary pattern in relation to child externalizing difficulties.
In sum, the evidence so far relating to the comparison of association strengths between healthy and unhealthy diets is equivocal. Perhaps the most plausible reason for this is due to heterogeneous methodological approaches related to the dietary definitions and statistical analyses chosen, which should serve as an encouragement to develop standardized methodological approaches within the nutritional epidemiological field.
Strengths and limitations
The strengths of this study include a large sample size, validation of dietary indices used and use of IPW to account for selection bias. We use robust statistical methods that to our knowledge have not been previously used within the nutritional epidemiology field. Also, investigations of associations between prenatal diet quality and child ADHD diagnosis has not previously been investigated.
One limitation of this study is that self-reported information, particularly in relation to aspects of health, introduces many challenges [
75‐
77], and collecting data with FFQ’s has generated much criticism [
78]. However, the maternal FFQ utilized in this study has been extensively validated and was explicitly developed for the target population [
34]. Moreover, using a composite measure of overall diet quality is more robust than looking at e.g. estimations of single nutrients with regards to misreporting, and it is a recommended method for investigating diet-disease relationships [
79].
Another limitation is that the MoBa is a selected group of participants and not representative of the whole population of Norwegian mothers and children. On average, MoBa mothers are older and more educated than the general pregnant population in Norway [
80], and both these factors are related to better prenatal diet quality and lower levels of child difficulties. However, as we used IPW for maternal age, education and child parity in our analyses, the results might be generalized with some caution, to populations similar to MoBa’s source population.
There are some limitations related to the information on ADHD diagnosis from the NPR. The youngest children in MoBa were born in 2009 and we have information on ADHD diagnosis in MoBa up to 2017 in our data file, hence the youngest children are 8 years old in the NPR sample. As children, especially girls, might receive an ADHD diagnosis at an age older than 8, there might be some false negatives in the sample, which has been found in the comparative pregnancy cohort in Denmark [
81]. In addition, a recent study investigating the practice of diagnosing ADHD in children in Norway found that only about half of the diagnoses were properly documented in the medical records, with inadequate differential diagnostic assessment being the main reason for unsatisfactory documentation for the ADHD diagnosis [
82].
Conclusion
In this paper we found that higher overall maternal diet quality and lower consumption of ultra-processed foods in pregnancy were associated with lower child ADHD symptom scores, but the association estimates were small. For child ADHD diagnosis, lower risk was found only for higher overall maternal diet quality measure. We found no associations of child diet quality at 3 years with either outcome.
We emphasize that no inferences regarding causation should be made based on these results, as unmeasured confounding could contribute to the observed associations. Also, as this is the first study investigating maternal and child diet quality in relation to both child ADHD symptom score and ADHD diagnosis, it is too early to draw firm conclusions about the associations we found. Instead we encourage more research conducted on this topic, with robust methodological approaches related to study design, variable definitions and statistical analyses, which will allow for better identification of possible causal relationships.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.