Introduction
Ever since the influential work of Lapouse and Monk [
1], the first true child psychiatric epidemiological study published in 1958, epidemiological strategies are inseparably intertwined with research and clinical practice in child and adolescent psychiatry. Although psychiatric epidemiology is often associated with the study of the frequency of disorders in the general population, epidemiological approaches have been used to answer questions pertaining to aetiology, diagnostic assessment, treatment and prevention of child psychopathology. Epidemiological approaches to study child psychopathology have developed rapidly from descriptive, cross-sectional studies in the 1960s, often using ad hoc procedures to assess problem behaviours reported by one informant (e.g. parents or teachers), to the current large-scale prospective cohorts integrating biological measures and multi-informant behavioural assessments to unravel aetiological mechanisms.
We will present (1) a brief and selective overview of landmark epidemiological general population-based studies that have influenced child psychiatry; (2) a more detailed overview of one currently conducted large-scale longitudinal cohort study that started pre-birth, the Generation R Study; (3) a discussion of methodological issues relevant to longitudinal epidemiological research and (4) suggestions for future studies that may move child psychiatry further.
The Generation R Study
The Generation R Study (“R” for Rotterdam) is a population-based cohort of children who are followed from foetal life forwards. This multi-ethnic cohort consists of 9,778 pregnant women and their live-born children. A subgroup of 1,106 women and their children received in-depth assessments. The overall aim of the study was to identify early determinants of children’s growth, development and health, including behavioural, cognitive and social development. A specific aim was to test the foetal origins hypothesis stating that disproportionate foetal growth programs later disease [
56].
Among its many measures, we believe that Generation R, compared to other cohort studies, stands out for the following: (1) the use of three prenatal ultrasound measures of the foetal head, brain (including ventricles) and other body parts, enabling the computation of prenatal growth curves; (2) its observational measures, including observations of the home environment, parent–child attachment relationship, parental sensitivity, the child’s compliance, temperament, emotion recognition and cheating; (3) neurodevelopmental measures, including postnatal 3D brain ultrasound, motor assessment, executive functions, IQ and language, and brain MRI (structural, DTI and resting state fMRI) and (4) reports on child problem behaviours by multiple informants: fathers, mothers, teachers, child self-reports and peer nominations [
52,
53].
We highlight three research themes that, to our opinion, have generated new insights: (1) associations between maternal prenatal exposures and foetal growth; (2) associations between foetal growth and postnatal child development and (3) associations between social disadvantage and foetal growth (see, supplementary Table 2 for an overview of major findings, recommendations and methodological challenges of the prenatal and early childhood neurodevelopmental and behavioural studies in Generation R; supplementary Table 3 gives an overview of references of published prenatal and early childhood neurodevelopmental and behavioural studies in Generation R).
Prenatal exposures and foetal growth
The repeated prenatal ultrasounds made it possible to study the impact of maternal exposures on trajectories of foetal head growth, which reflects early neurodevelopment. Similarly, indictors of foetal cerebral blood flow have been assessed by ultrasound to study neurodevelopment. Maternal smoking, cannabis or SSRI use, low folate during pregnancy and prenatal maternal depression, anxiety and family stress were all associated with less foetal head growth, with relatively strong effect estimates observed for maternal smoking, SSRI and cannabis use [
57‐
59]. A finding that stood out was the substantial negative effect of (subclinical) hypothyroidism on foetal growth and birth outcomes [
60]. We concluded that numerous intrauterine exposures are related to foetal growth from the first to third trimester.
Foetal growth and postnatal child development
Low birth weight is associated with several psychiatric problems, including depression and hyperactivity, and poor cognitive functioning [
61,
62]. However, birth weight is an end point measure of foetal growth and does not inform us about variations in intrauterine growth. Therefore, we tested whether prenatal exposures impacted on preschool child behaviour and cognition and to what extent this was mediated by variations in foetal growth trajectories. Although we found that intrauterine head growth was associated with delayed early motor development, there was little indication that intrauterine growth was associated with problem behaviour, temperamental difficulties or cognition [
63,
64]. Also, we found that maternal exposures negatively affecting intrauterine growth, such as exposures to nicotine or cannabis, had little, if any, impact on later child problem behaviour if carefully controlled for confounding [
63]. If prenatal environmental exposures were associated with later problem behaviours in the preschool period, as was the case for family stress or parental psychopathology, there was no mediation by foetal growth, suggesting that much of the observed intrauterine association was due to genetic confounding or spill over of parental behaviour from the prenatal to the postnatal period [
65]. In conclusion, these findings support the view that several negative effects of prenatal environmental exposures can be compensated later in life. Hence, detailed observations of foetal growth in the Generation R cohort did not support the Barker hypothesis. Of course, we do not know yet what the impact of these environmental exposures is on later functioning at school age and in adolescence.
Social disadvantage and foetal growth
The impact of social disadvantage can be detected from childhood onwards. We found compelling evidence that socioeconomic differences can already be found in the intrauterine period and that this extends into the child’s postnatal development. Low maternal education was not only associated with premature birth and low birth weight, but also with less foetal growth. Importantly, social differences were detectable in foetal brain growth more than in that of peripheral and abdominal tissues [
66]. The association between poor foetal growth and child problem behaviours could largely be explained by parental psychosocial and maternal lifestyle characteristics [
63]. Socioeconomic inequalities could also be documented for early temperamental traits and problem behaviour in young children [
67]. This indicates that health inequalities accumulate in disadvantaged families from the intrauterine period of children’s life onwards.
So far, most Generation R studies have shown that individual differences arise from a large number of causal factors, with each contributing a relatively small effect. This cumulative risk model suggests that the additive contribution of genetic, perinatal and environmental risks puts the child at a progressively greater risk, despite the small impact that any single factor is likely to have.
Suggestions for future studies
Epidemiological studies have been successful in describing the frequency and course of child psychiatric problems. Aetiological research clearly demonstrated the strong associations between social disadvantage and child maltreatment with behavioural and emotional development. In contrast, the high expectations that biological factors can be used to better explain, diagnose or predict child psychiatric problems have not been met.
Future risk factor epidemiology in child psychiatry will undoubtedly follow the lead of cardiovascular and cancer research. After the hype of candidate gene, cGXE and GWAS studies, child psychiatrists are already focussing on epigenetic, microbiomic and metabolomic studies. Yet, child psychiatry has hardly harvested the fruit on the GWAS tree. So far, published studies of common child psychiatric problems have been hopelessly underpowered. If schizophrenia researchers need to include more than 30,000 affected individuals to find several genetic loci, child psychiatrists must certainly include equal numbers of children with disorders or, because large mono-site samples may not be feasible, pool cohort studies exceeding 100,000 children to find associations between genetic variants and common psychiatric disorders or traits. This is not so much because child psychiatric disorders are less heritable, but mainly because disorders are less specific, not precisely measured and subject to developmental changes. Pooling across multiple sites, either within or across nationalities, should ideally comprise information on individuals that were sampled and assessed according to the same protocol as was done, for example, in the IMAGEN study [
81]. Pooling of samples from ongoing studies that use different methodologies has the disadvantage that only the limited number of confounders, which were measured uniformly across studies, can be taken into account. Of course, natural experiments, such as the Romanian adoption studies [
82], or studies using high-risk samples, such as the Mannheim Study of Children at Risk [
83] can also be useful to study the aetiology of child psychopathology, but fall outside the scope of this selected review.
The future epigenetic, metabolomics and microbiomic studies should adhere to the lessons learnt from the (non-psychiatric) GWAS studies. The scientific rigor with multiple hypothesis testing and immediate replication must be adapted and applied before child psychiatrists embrace this new research opportunity. In the meantime, child psychiatry may devote some attention to new environmental risks provided by extensive media use and chemicals. Phthalates and bisphosphonates are toxins with a presumably neurodevelopmental effect. If not cohort researchers, who is in the position to assess the impact of these compounds?.
Traditionally, child psychiatric researchers have relied on continuous phenotype measures, both in clinical practice and research. Likewise, child psychiatrists have recommended a multiple informant approach to assess child psychopathology. Although implemented in clinical practice, researchers do not systematically adhere to this advice. In particular in biological studies such as neuroimaging or genetic research, few examples of multiple informant assessments can be found. Future studies can improve precision and validity of results by complementing mother reports with father, child self, peer or teacher reports of child problem behaviours.
Clinical cohorts have successfully incorporated neuroimaging techniques in both cross-sectional and longitudinal design. Notably, a maturational delay was demonstrated in children with ADHD and abnormal neurodevelopmental processes in adolescents with early-onset schizophrenia [
84,
85]. These findings cannot unravel the direction of the association between disorder and brain development. Population neuroscience is needed to demonstrate whether brain alterations precede the onset of psychiatric problems. This implies that large-scale imaging studies in very young children from the general population are performed.
In summary, future research in child psychiatry will most probably continue to employ sophisticated epidemiological approaches. Multiple imputations, immediate replication, models of causal inference, repeated measures and multiple informants are increasingly standard practice. Yet to apply genetics, neuroscience or other molecular research to better understand how the brain produces maladaptive behaviour, we need to conduct more ambitious large-scale child psychiatric cohorts. Progress is booked only if we integrate the basic science with more rigorous epidemiological designs.