Final Report Summary - ACTION (Aggression in Children: Unraveling gene-environment interplay to inform Treatment and InterventiON strateg)
Executive Summary:
Childhood aggression can be described in many ways, by multiple people who are familiar with a child and by many instruments and methods. The Action consortium investigated causes and consequences of aggressive behavior in childhood, as assessed from multiple perspectives, in order to improve the understanding of the genetic and non-genetic etiology of aggression in children and to inform on the development of prevention and treatment strategies. Action brought together data from twin cohorts and genotyped cohorts to estimate and identify the causes of variation in aggression, to disentangle (epi)genetic and environmental effects, to inform on comorbid disorders, to investigate biomarkers and to gain insight in the predictive power of pediatric aggression for later outcomes.
We showed that aggressive behavior in childhood is associated with negative behavioral outcomes and that that it predicts emotional and cognitive problems. We established that early detection of aggressive problems and access to treatment results in significantly lower crime records and higher average income. We also showed that across development and environmental context, aggressive problems do not occur in isolation: the majority of children who are aggressive also suffer from attention problems, anxiety and depression. There is a robust negative association between aggression and academic performance in compulsory education. The large comorbidity of aggression with other childhood psychopathology was also indicated by clinicians, social workers, parents and teachers of children as one of the major difficulties in treating this group.
To obtain an understanding of why children differ from each in aggression, ACTION established an international consortium among partners with access to large childhood cohorts from Finland, Sweden, UK, the Netherlands, Italy and Australia. The longitudinal data established the substantial heritability of aggression across age and gender. The stability of aggression across age is large and due to the constant genetic influences on aggression. Next, Action extended the consortium to include many research groups from around the globe, in meta-analysis projects that focus on identification of genetic markers and differentially methylated regions of aggression and one of its major comorbid conditions, i.e. attention problems.
The genome-wide epigenetic results based on a meta-analysis of 14,434 peripheral blood samples across all ages and 2425 cord blood samples indicated 48 top CpGs for which we performed enrichment analyses. The top ten most strongly enriched traits indicate that CpGs associated with aggressive behavior overlap with CpGs previously associated with smoking, maternal smoking, and other chemical exposures (e.g. perinatal exposure to perinatal polychlorinated biphenyls (PCBs) and polychlorinated dibenzofurans (PCDFs)). Further overlap includes CpGs associated with alcohol use, cognitive function, educational attainment, ageing, and metabolic traits.
The first large genetic association meta-analysis (GWAMA) of childhood aggression identified two genes that are significantly associated with AGG in a gene-based analysis: PCDH7 (protocadherin 7) on chromosome 4p15.1 which has the highest expression in the brain and spleen and ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase 3) located on 1p34.1 which is most highly expressed in testis and fat tissue. There was significant heritability based on single nucleotide polymorphisms. Genetic correlations were high and positive for a series of disorders and behaviors including ADHD, Bipolar disorder, Loneliness, Major depression , Risk tolerance, and Smoking. Genetic correlations were high and negative for educational attainment and age of first child.
Project Context and Objectives:
Within Action, we worked on a set of interrelated objectives: Within Clinical Epidemiology we aimed to 1) gain insight in the treatment and diagnostic practices in Europe and identify the critical needs for stakeholders; 2) quantify and identify the long-term societal outcomes of childhood aggressive behavior; and 3) develop long-term positive and negative trajectories from early childhood to adulthood for subtypes of aggression. In Genetic Epidemiology, we aimed to 1) unravel the genetic longitudinal architecture of aggression and estimate heritability; 2) resolve patterns of comorbidity of aggression and emotional and behavioral problems; 3) detect genomic regions for aggression in children in genome-wide association (GWA) studies; and 4) identify differentially methylated regions related to aggression, also in a genome-wide approach. In Gene-Environment Interplay we focused on environment in order to: 1) identify environmental risks that have effects on the development of aggression independent of genetics; 2) identify environmental effects on the development of aggression that depend on genetic propensities (GxE interaction); 3) identify genetically driven experiences associated with aggression in which children select, modify or create environments correlated with their genetic propensities (GE correlation); 4) apply methods for SNP heritability and polygenetic predictors; and 5) extend analyses of epigenetics to investigate GE interplay. For the work on Metabolomics & Biomarkers we collected buccal swabs and urine samples in children from clinical and population settings. Action aimed to 1) identify biomarkers to diagnose aggression; 2) identify biomarkers to differentiate between sub-groups of aggression; and to 3) identify treatment options.
The work realized in these four large epidemiologically oriented approaches informed on treatment and prevention strategies and lead to 1) development of an evidence-based model on effectiveness of existing preventative and treatment strategies; 2) integration of insights in a comprehensive theoretical framework; 3) development of risk assessment charts and guidelines to improve clinical decision making; and the assessment of this framework and risk assessment charts with respect to ethical, legal and social implications (ELSI). Within our work package Dissemination and stakeholder interaction we collaborated to 1) provide a recognizable and global project identity; 2) update participants and participating agencies; 3) inform stakeholders and general public; 4) inform the scientific community about the project; and 5) create a list of stakeholders and end-user groups.
In our clinical epidemiology study of experts and clinicians from 24 countries of practices in Europe, experts supported developing European guidelines and emphasized a need to develop Evidence Based Methods. Clinicians emphasized limited usefulness of guidelines for clinical practice. Therapeutic alliance, focus on positives, improving parent-child interactions, and collaboration with schools and teachers were stressed to improve prevention and treatment of severe behavioral problems. Comorbidity was identified as a major theme. Comorbidity was the focus of one of two of our large empirical study across the childhood cohorts in ACTION. We showed that co-occurrence of aggression and other behavioral and emotional problems is substantial, and that this pattern strongly emerges independent the person who rates the child’s behavior (parent or teacher). There is an especially high degree of co-occurrence of aggression with attention problems, rule breaking, oppositional behavior, and substantial associations are also seen with depression, anxiety, and autism. The ACTION website has a set of interactive tools www.action-euproject.eu/ComorbidityChildAggression and
http://www.action-euproject.eu/content/new-interactive-tool-released showing the comorbidities of aggression with a wide range of childhood problems, across ages, raters and instruments.
A record linkage study in Sweden of long-term societal outcomes of childhood aggression revealed how long-term positive and negative trajectories from early childhood to adulthood develop for aggression subtypes; how early detection of aggressive problems and access to treatment results in significantly lower crime records and higher average income. Suicide risk might also decrease.
A systematic literature review of effectiveness of prevention and intervention described effect sizes across treatments and moderator effects. For prevention and intervention, effects were small or medium. For moderators, effects were mixed. Moderators with a positive effect on treatment effectiveness for childhood aggression were pre-test levels of aggression, individual implementation, and parental involvement. Future research should distinguish between targets of treatment, focus on individual differences between children in treatment effectiveness and parental influences. A second systemic review summarized the current knowledge on the genetic basis of human aggression. As part of this project, we introduced an automated tool for the selection of literature based on supervised machine learning. This review summarizes all heritability and genetic association studies of aggression and concluded that power gains for GWASs on aggression need to be realized by increasing sample sizes, by combining data from different cohorts and by meta-analyzing genetically related outcomes.
Genetics plays an important role in explaining the individual differences in childhood aggression with heritability estimates of >50% and much smaller contributions of cultural inheritance and shared environment and genetic factors also are the main drivers of stability in aggressive behavior. A phenotype reference panel was realized, in which aggression was assessed on the same children by multiple instruments as used in the ACTION cohorts. We realized a survey study that included the CBCL (used in Holland and Sweden), the SDQ (used in UK), and the A-TAC (used in Sweden) for 2508 twins. Our results demonstrated heritability and high genetic overlap, suggesting that the underlying construct of childhood AGG is consistent across measures. This finding holds promise for gene finding studies.
To identify genetic variants, two large genetic association meta-analyses (GWAMA) for Aggression and ADHD were finalized, for which new meta-analytic approaches were developed. The GWAMAs include data from 30 cohorts, extending the Action consortium significantly to other childhood cohorts, with over 500.000 data points from multiple raters. Two genes are significantly associated with AGG in the gene-based analysis: PCDH7 (protocadherin 7) located on 4p15.1 which has the highest expression in the brain and spleen and ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase 3) located on 1p34.1. It is most highly expressed in testis and fat tissue. While we find no genome-wide significant SNPs in the overall GWAMA, we find significant SNP heritability and genetic correlations for a wide range of outcomes: childhood AGG is substantially correlated with ADHD and adult psychiatric disorders, including Bipolar Disorder, Schizophrenia , and Major depression. Substantial correlations were also seen for loneliness, risk tolerance, and smoking. Genetic correlations were high and negative for educational attainment and age of first child. The GWAMAs provide important information for the broader psychiatric genetics community. Action will make the meta-analysis results available, so that polygenetic scores (PGS) can be obtained for prediction in independent cohorts. Equally important are results for research on modification of genetic influences by environment, either through considering the totality of such influences by modeling of genetic relatedness matrices, or though considering the interaction with PGS and environment.
To identify differentially methylated regions we also were able to contact and include multiple international cohorts. Two EWAS projects that carried out meta-analysis of internationally available data were completed: one included 3 cohorts with genome-wide epigenetic data and ADHD. Findings point at new candidate loci involved in immune and neuronal function. The second EWAS includes all cohorts with aggression phenotypes and genome-wide epigenetic data, again extending the consortium far beyond its original members. This EWAMA includes 20 cohorts with data for 14,438 peripheral blood samples (mean age at blood sampling from 4 to 70 years) and 2195 cord blood samples. The mean age at phenotypic assessment is 6 to 72 years and aggressive or conduct problems were assessed by multiple instruments. In 48 top CpGs from the meta-analysis we performed enrichment analyses in the EWAS atlas against all reported associations with diseases and environmental exposures. The top ten most strongly enriched traits indicate that CpGs associated with aggressive behavior overlap with CpGs previously associated with smoking (37 CpGs; corresponding to 77% of aggression-associated CpGs and 0.3% of CpGs that were previously associated with smoking), with other smoking traits (e.g. maternal smoking), other chemical exposures (e.g. perinatal exposure to perinatal polychlorinated biphenyls (PCBs) and polychlorinated dibenzofurans (PCDFs)). Further overlap includes CpGs associated with alcohol use, cognitive function, educational attainment, ageing, and metabolic traits. These results confirm the important role of that exposure to nicotine that was also established in the analyses of the twin cohort data.
Our study of environment and interplay of genes and environment shows that perinatal factors explained up to 5% of the variance in adolescent aggression. Maternal smoking during pregnancy predicted aggression at all ages independent of other perinatal and demographic factors. A multidimensional risk score based on home and school risks predicted 10% of adolescent aggression. The multidimensional risk factors of aggression spanned multiple domains of adolescents’ lives. Integrating environmental and polygenic predictors and assessing their predictive accuracy indicate that genotype-environment interaction and correlations are of very small effect size.
The study of biomarkers in urine has been done in a technical (n=20), a biochemical pilot (n=218) and in the main biomarker study (n= 1142 selected (on high and low aggression) twins and 186 clinical cases. Assessments were done for Creatinine, Density, Neopterin, C-peptide, Substance P, Oxidized DNA and RNA and on metabolomics platforms for Amines and Organic Acids platforms. We began with a review of the existing literature and identified 4 classes of potential biomarkers for human aggression: 1) lipids, 2) inflammation markers, 3) neurotransmitters and 4) hormones. With the work performed in Action we find suggestive evidence for the role of inflammation markers, lipids and amino acids involved in neuro-transmitter pathways in aggression. In children we identified metabolites of the cysteine metabolism pathway, involved in oxidative stress, as among the most promising aggression biomarkers. In a project in adults we found cytokines to be marginally significantly related to aggression differences in discordant MZ pairs and LDL cholesterol levels to be lower for the high scoring MZ twins as compared to their co-twin scoring low on aggression. Finally, the top 25% most associated metabolites in ACTION’s Main Study included multiple (essential) amino acids which are involved in neurotransmitter pathways, or may act as neurotransmitters themselves, such as the metabolites included in the cysteine metabolism pathway.
Research on the biological and genetic components of aggression needs to be balanced by attention to the context and environment in which children develop. ACTION contributes to a better understanding of aggression by examining children’s aggressive behavior in the context in which it develops. A child who is aggressive in the school context requires different treatment than one who is aggressive in the context of chronic stress associated with parental psychopathology. The type of intervention that is adaptive for children exposed to long-lasting economic hardship and socio-economic disadvantage may differ from youth who live in more economically advantaged environments. Polygenic scores (PGS ) based on genetic variants from the discovery GWA results can be applied, together with the results from the methylation, biomarker, environmental context analyses to build models predicting which children have a high risk for persistent aggression. This model can be used in future for stratification of high and low risk groups to investigate whether and which more intense treatments improve outcomes for high risk children.
The ACTION website is continuously updated and now also includes the Standard Operating Procedures for collaborative projects (http://www.action-euproject.eu/content/data-protocols ), publications, videos, interviews with (young) researchers and newsletters.
Our Scientific Advisory Board concluded “ACTION has delivered a thorough foundation for ongoing and future basic studies into childhood aggression especially by producing extensive and complete overviews on clinical practices and on genetic influences on aggression. The consortium has invested in innovative methodology for literature mining and established a large collaboration of cohort and clinical studies into childhood aggression in Finland, Sweden, UK, The Netherlands, Italy and Australia. A diversity of highly relevant aspects of this topic was studied in these large datasets such as the correlated morbidities that coincide with aggression in children, parental factors, the stability of genetic influences over time (very stable in longitudinal studies), genetic overlap of aggression with other disorders (such as ADHD) and longstanding questions such as the role of maternal smoking during pregnancy. Scientifically the project is a great success with 90 papers published and many more expected to come in the next year on the basis of the novel data collected. Data in these cohorts have been harmonized and brought together for basic studies into genetics and early and later environmental factors pivotal to the development of childhood aggression. From a public health point of view, the EU was right in supporting this multidisciplinary project which is and will be producing robust evidence for the insights generated”.
Project Results:
Below, we report for our Action project [Aggression in Children: Unraveling gene-environment interplay to inform Treatment and InterventiON strategies] the main outcomes for all objectives, as initially stated in our application, to address questions regarding paediatric conduct disorders and aggressive behavior. Our Action project had as its broad aim to improve the understanding of the genetic and non-genetic etiology of aggression in children, with the intention to inform the development of novel prevention and treatment strategies. We set out to unravel in large twin and genotyped cohorts the causes of variation in aggression, disentangling (epi)genetic and environmental effects and their interplay with a focus on critical developmental periods, gender, and comorbid disorders; to examine metabolomics profiles of aggressive behavior and gain insight regarding the predictive power of pediatric aggression for adult outcome variables.
Below, the results for each of the objectives of the project are summarized. These objectives were part of a series of interrelated work packages, that had their focus on Clinical Epidemiology, Genetic Epidemiology, Gene-environment interplay, Biomarkers, Societal implications and Dissemination.
Results for Objectives from our work package Clinical Epidemiology.
Objective 2.1: Select for sample collection for epigenetic and biomarkers studies children (aged 6-12 year) from clinical setting and collect buccal swabs and urine
Sample collection was realized in 189 children. Based on our empirical experiences, we addressed the difficulty of including participants from clinical settings. In projects such as ours, researchers can be faced with decreases in the number of children admitted to clinics, because of reorganization of child and adolescent psychiatry policies, such as happened in the Netherlands. Related to this, the relative number of children with acute mental health problems increased and due to the reorganization of Dutch child and adolescent psychiatry and severe budget cuts, institutions had to decline participation. We have analyzed the whole process by approaching parents of children under these conditions. Based on a quantitative analysis of (non-)participants, the following changes in the recruitment procedure have been implemented: the content of telephonic interviews was adjusted to offer options that make participation easier, e.g. the possibility of a researcher making a house visit instead of the child having to visit the clinic or the possibility that a researcher instead of a parent collects the urine and buccal swaps of children who are spending the night at a clinic. A more personal approach was adopted to explain the study. A pilot was done in which parents who indicated interest in participating during the first contact, receive study details in a face-to face setting. We consider this better than mailing study folders and informed consent forms. Reasons given by parents largely relate to the stress experienced when entering a mental health facility. They feel too exhausted to invest in research participation, although they consider this of great importance.
Objective 2.2: Gain insight into the current treatment practices (guidelines, interventions and working components) for severe behavioral problems (SBPs) in children in Europe and identify the critical needs for stakeholders.
This objective was reached through conducting qualitative interviews in Europe to investigate the availability and utility of clinical guidelines for diagnostics and treatment of childhood aggression, according to academic experts’ perceptions (study 2.2.1); the diagnostics and treatment practices, and critical needs, according to mental health clinicians (study 2.2.2); and the availability of prevention and intervention programs for childhood aggression, according to academic experts’ perspectives (study 2.2.3). An additional study was compiled which reported on the awareness and perceptions of clinical guidelines, according to mental health clinicians, and overlapped this knowledge to that from experts’ reports on guidelines (study 2.2.4). Due to high variability in characterizing and diagnosing aggression in childhood across Europe, the umbrella concept of Severe Behavioral Problems (SBPs) was introduced. SBPs include frequent oppositional, aggressive and destructive behaviors, occurring across settings and impairing daily functioning.
2.2.1 Study Experts guidelines: ‘Awareness and perceptions of clinical guidelines for the diagnostics and treatment of severe behavioral problems in children across Europe: A qualitative survey with academic experts.
This study highlights the current successes and challenges perceived by experts around Europe associated with guidelines and documents for SBPs in children. Additionally, it establishes experts’ consensus for the need to develop official guidelines better tailored to clinical practice, creating an appropriate momentum for a transition towards European clinical guidelines for this population.
Some of the findings on guidelines availability have also been incorporated into a follow-up study and compared against mental health clinicians’ awareness of guidelines (see study 2.2.4. below).
2.2.2: Study Clinical practices: ‘Inventory of mental health clinicians' practices and critical needs for managing Severe Behaviour Problems (SBPs) in children across Europe’, according to mental health clinicians’ perspectives’
This project gathered feedback from mental health clinicians about diagnostic and treatment tools used and available in practice. Our recruitment reached 500 clinicians from 30 European countries. We received complete responses from 138 clinicians from 24 European countries, who reported through an online semi-structured questionnaire developed by the authors. Most clinicians had a medical training background (i.e. psychiatrists and psychiatrists in training; 53%), followed by psychotherapy and psychology. Clinicians perceived high severity of symptoms, comorbidity (e.g. ADHD, depression), poor daily functioning, and deficient family environment as risk factors indicating need for specialist treatment. Clinicians reported primarily using cognitive-behavior, pharmacotherapy and behavioral management programs as in clinical practice. Other specific programs used included the Incredible Years, Problem-Solving Skills Training, Anger Coping Program, VIG, and Triple-P. Clinicians also commented on critical needs in current practice. The main challenges include poor multi-agency collaboration, stigma and lack of awareness of SBPS, lack of resources, little support for teachers, and poor responsibility taking across systems. Clinicians’ main focus for future improvements included early detection and prevention (23%), multi-agency collaboration (21%) and increasing knowledge of SBPs (16%), although many clinicians had a pessimistic vision of a future (18%) with little funding and no prioritization of SBPs (12%). These findings highlight existing strategies and gaps in service provision for SBPs according to clinicians’ perspectives informing further action for improvement.
2.2.3. Study: Clinical practices: ‘Availability of prevention and intervention programs for severe behavioral problems in children across Europe, according to academic experts’ perspectives’
One interest of this study focused on in-depth investigation into experts’ opinions on accessibility of services and critical needs related to these programs. Numerous programs were mentioned, ranging from CBT-based interventions, Parent Training programs, or multi-systemic therapy to social services involvement and placement in residential care. Critical needs included improving implementation of these evidence-based programs by increasing their accessibility and increasing staff training, more evidence-based treatments, particularly behavioral-based parenting programs, more standardization of treatment, or wider dissemination. Differences across Europe were also identified with regards to how accessible services are for SBPs. Main challenges revolved around high demand over capacity. Overall, these findings provide a summary of available prevention and intervention programs across Europe, emphasizing existing gaps in service provision.
2.2.4 Study: Clinicians guidelines: ‘Perspectives on clinical guidelines for severe behavioural problems in children across Europe. A qualitative study with mental health clinicians’
We found that under half of the clinicians reported being unaware of any guidelines. Of these, 37% represented countries where guidelines were available according to experts. The remaining half of clinicians who were aware of guidelines on average reported being moderately familiar with their content, perceiving them as moderately useful and using them some of the time. Additionally, 60% clinicians agreed that SBPs guidelines need to be developed, as these would create a shared knowledge base for diagnosis, prevention and treatment. Guideline improvements included taking a multifactorial approach, creating specific case recommendations, encouraging multi-agency work. The need for additional funding, adaptation to local practices, and personalization to individual cases was also highlighted. These critical needs matched those highlighted by experts in study 2.2.1. Overall, the findings revealed that the modest familiarity with and use of guidelines amongst practitioners may highlight guidelines poor fitness-for-purpose, or, alternatively, an underlying confusion around the meaning and purpose of guidelines. Moving forward, efforts should be directed at disseminating clearer definitions of and instructions on using guidelines, and addressing current needs identified in this study.
Objective 2.3: Quantifying and identifying the long-term societal outcomes of childhood aggressive behavior.
Findings from objective 2.2 have substantially guided our activities in this objective.
2.3.1 Study Children with early-onset disruptive behavior: parental mental disorders predict poor psychosocial functioning in adolescence
The study provided novel evidence that parental MD puts 9-year-olds with DB at risk for negative outcomes in adolescence, both in terms of antisocial behavior and other outcomes, like school performance and truancy. Additionally, paternal MD was a better predictor than maternal MD, regardless of child DB at age 9, suggesting that fathers should be given increased attention in future research. Treatment-as-usual of children with DB could be augmented with additional screening and, if necessary, treatment of mental health problems in their parents.
2.3.2 Study The relation between the age at diagnosis of problem behaviors related to aggression and distal outcomes in Swedish children.
Severe childhood aggressive behaviors are known to predict negative outcomes later in life; however, little is known about the effect of when in childhood aggression problems are diagnosed. We analyzed data from the population-based Swedish Data Registries, which include data on all children formally diagnosed by the Swedish medical system with a wide range of aggression problems between ages 8 and 18 (N = 5,816) during the years 1987 to 2013, along with a matched control. Time-to-event analyses investigated whether the age at time of diagnosis affects later life outcomes while controlling for relevant confounders. Results show that for both boys and girls those with a later diagnosis had lower average incomes and a higher probability of having a criminal record. The effect on suicide attempts was not significant after correcting for multiple testing. Grade score was not significantly affected. The results warrant further research concerning advantage of earlier diagnoses and generalizability beyond the Swedish population.
2.3.3 Study Assessing Psychopathic Traits: Validation Study of The Child Problematic Traits Inventory and the Instrument of Reactive and Proactive Aggression in a Clinical Sample
Given the emphasis on empathy deficits, a clinical relevant focus related to empathy includes callous-unemotional (CU) traits. We ascertained whether CU measures are reliable and valid in clinic-referred youth and tested the psychometric properties of the Child Problematic Traits Inventory (CPTI) in clinic-referred children (ages 6-13 years). Teachers (N = 159) and parents (N = 173) completed the CPTI. Confirmatory factor analyses supported the CPTI’s 3-factor structure when teachers and parents rated the 28 CPTI items. Teacher- and parent-reported CPTI scores showed the expected relations with external correlates (e.g. conduct problems, and proactive aggression). Crucially, the validity of the CPTI scores was also supported across informants and across methods (i.e. regardless if a questionnaire or a diagnostic interview was used to measure external correlates). The CPTI holds promise as an assessment tool for assessing psychopathic traits in clinic-referred children. The parent- and teacher-reported Instrument of Reactive and Proactive Aggression (IRPA), offers another measure of severity of behavioral problems, and distinguishes between functions (reactive and proactive) and forms (physical/covert/verbal) of aggressive behavior. The IRPA will also be used to study subtypes of aggressive behavior within the clinical sample that has participated in the epigenetic and biomarker studies.
Objective 2.4:Develop long-term positive and negative trajectories from (early) childhood to late adulthood for subtypes and heterogeneity of aggression.
2.4.1. Study Can Mental Health Screening of Detained Male Juveniles Provide Evidence of Likelihood of Past and Future Violent Criminal Behavior?
In data from youth detention centers from the Netherlands we explored whether brief mental health screening tools are of value for alerting staff to a detained youth’s past offending or potential future violent offending. Boys (n = 1559; Mean age = 16.65) completed screening questionnaires that tap conduct problems, risk factors for aggression (e.g. anger-irritability), and mental health problems (e.g. depressed/anxiety). Official records were collected to index past and future violent offending. The small number of relations and their small effect sizes suggest little likelihood that screening for conduct problems, risk factors of aggression (e.g. anger-irritability) and mental health problems in boys who are detained in the Netherlands offers any potential for identifying past and future violent arrests.
2.4.2. Childhood aggression predicts externalizing but also internalizing adverse outcomes in emerging adulthood
In this longitudinal study, we examined the association between childhood aggression (based on parent-report; n = 18,649) with adverse outcomes in emerging adulthood identified in Swedish national registers, including rates of clinically diagnosed mental disorders, prescription of anti-depressant/sedative medications, suicide attempts, alcohol/substance use disorders, criminal convictions, failure to reach eligibility for high school, and social welfare support. We examined if childhood aggression predicted internalizing problems after adjusting for its overlap with externalizing problems and examined the genetic and environmental contributions to these associations using twin modeling. We found that childhood aggression predicted all outcomes in emerging adulthood. Childhood aggression was associated not only with a general problem factor and the independent externalizing factor, but also with the independent internalizing factor. The twin analyses showed that genetics contributed to the association between childhood aggression and all three outcome factors, whereas the shared environment seemed to be more important for the association between childhood aggression and the externalizing. The non-shared environment contributed marginally to all associations. These results indicate that childhood aggression is an important indicator of a host of adversities in emerging adulthood. Prevention and intervention of aggression at an early age may pay large dividends above and beyond aggressive behavior as it may also reduce anxiety, depression and several functional outcomes.
2.4.3. Harmonizing behavioral outcomes across studies, raters, and countries: an application to the genetic analysis of aggression in the ACTION Consortium
To include considerations of social and interaction effects requires large datasets with harmonized outcome measures. Here, we made use of a reference panel for phenotype data to harmonize multiple aggression measures in school-children to jointly analyze data from five large twin cohorts on 86,612 children (42,468 twin pairs) from five European twin cohorts. A bi-factor integration model in the integrative data analysis framework was developed to model aggression across studies while adjusting for rater, age, and sex. The harmonized aggression scores were analyzed to estimate contributions of genes, environment, and social interaction to aggression. The best-fitting model found a high level of overall heritability of aggression (~60%). Social interaction effects were significant in the opposite-sex twin pairs, and the model fit greatly improved when the social interaction effect of males on their female co-twin differed from the effect of females on their male co-twin. An aggressive female had a positive effect on male co-twin aggression, whereas more aggression in males had a negative influence on a female co-twin. The joint analysis with a large sample facilitated these new social interaction models for opposite-sex twins.
2.4.4. Effects of antipsychotics on future violence
Association Between Prescription of Major Psychotropic Medications and Violent Reoffending After Prison Release: This cohort study included all released prisoners in Sweden from July 1, 2005, to December 31, 2010, through linkage of population-based registers. Rates of violent reoffending during medicated periods were compared with rates during non-medicated periods in within-individual analyses. Follow-up ended December 31, 2013. Among released prisoners, rates of violent reoffending were lower during periods when individuals were dispensed antipsychotics, psychostimulants, and drugs for addictive disorders, compared with periods in which they were not dispensed these medications. Further research is needed to understand the causal nature of this association.
Second-generation antipsychotics and risk of violent crime in adolescents and young adults: Prescriptions of second-generation antipsychotics (SGAs) in young people have increased markedly during the last two decades, with a large proportion prescribed to those who exhibit disruptive behaviors and aggression. We investigated whether SGAs reduced the risk of violent crime in young people in a national cohort of individuals aged 15-24 years who have received at least one dispensed prescription of SGAs between 2006 and 2013 using data from Swedish national registers. Among young men received SGAs, rates of violent reoffending were lower during periods when individuals received SGAs compared with periods when they did not.
2.4.5. Ghrelin and aggressive behaviors
The relevance of ghrelin for human aggression per se as well as for aggression induced by alcohol was evaluated in a human genetic association study in young men (n=784) from the normal population assessed for anti-social behaviors. The study demonstrates that central ghrelin infusion, but not ghrelin administered systemically, increases aggression. Moreover aggressive behavior is decreased by pharmacological suppression of the growth hormone secretagogue receptor-1A (GHSR-1A). The genetic association study demonstrates that males carrying the Leu72Leu genotype of the pre-pro-ghrelin gene and displaying hazardous alcohol use are more aggressive when compared to the group carrying the Met-allele suggesting the identification of central ghrelin pathway as an important modulator in the onset of aggression.
Results for Objectives from our work package Genetic Epidemiology.
Objective 3.1: Select monozygotic twin pairs for sample collection for epigenetic and biomarkers studies
This objective was an essential step in the ACTION project. We collected buccal DNA and urine samples in a large group of twins concordant and discordant for aggressive behaviour. The buccal samples have been used to isolate DNA and to create methylation profiles. The methylation data have been cleaned and tested for association with aggression (Objective 3.5). The urine samples were collected for the biomarker research in WP5. Twin pairs from the large longitudinal database of the Netherlands Twin Register were selected. We targeted families with two control children (concordant control), two cases (concordant case) and families in which one of the twin pair is a control and the other a case (discordant pairs).
First-morning urine and buccal-cell DNA samples were collected across three phases of data collection: 1) the ‘technical pilot study’. This was a pilot which established feasibility and stability. Twins were unselected for phenotype information. First-morning urine and buccal-cell DNA was collected in twin pairs at two time points, with an interval of approximately two weeks. In total, 24 twin-children (12 twin pairs) were invited to participate. Permission for participation in the ‘technical pilot study’ was received from the parents of ten twin pairs (participation rate: 83.3%), 40% were girls. 2) the ‘biochemical study’. This was a large pilot to establish which biomarkers should be taken forward to the main study. We did collection of first-morning urine, buccal-cell DNA and assessed parent-reported aggression at the time of sample collection. Twin participants were invited as either aggression cases or controls, though complete twin pairs were always invited for participation. We invited 193 case-children, 395 control-children ( mean age of 9.8 years (SD = 1.6) 50.4% girls). The twins, of 293 families, belonged to 85 MZ twin pairs, 209 DZ twin pairs. Across all invitation we had a participation rate of 52% . We forwarded 88 participants to the next phase of the project, the ‘main study’. In total, the ‘biochemical study’ included 218 individuals of 6 to 11 years of age (mean age = 9.9; SD age = 1.6 51.2% females; 16 MZ, 93 DZ pairs). 3) the ‘main study’, collected data in MZ twin pairs concordant high concordant low on aggression and MZ twin pairs discordant on aggression. The mean age of 9.5 years (SD = 1.9) of the invited twins 49.6% were girls. The twins, from 1342 families, belonged to 1309 MZ twin pairs and 37 DZ twin pairs including 2 DOS twin pairs. Whilst BB3 aimed to include MZ twin pairs, DZ twin pairs were included for two reason: a) a selected MZ twin pair turned out to be a DZ twin pair after zygosity testing or b) the DZ twin pairs were siblings of a selected MZ twin pair and therefore also included in the study. Across all invitation batches we had a participation rate of 39.4% with 237 concordant low aggression twin pairs, 137 concordant high aggression twin pairs and 157 discordant twin pairs participating in the ‘main study’. In the ‘main study’ a total of 531 twin pairs agreed to participate, with the 44 twin pairs originally recruited as part of the ‘biochemical study’, and totals 575 twin pairs of 6 to 13 years of age (mean age = 9.5; SD age = 1.9 47.3% females) of which 255 were concordant low aggression twin pairs, 152 were concordant high aggression twin pairs and 168 were discordant twin pairs. There are 575 twin pairs (547 MZ and 30 DZ pairs).
The total response rate across all three phases of data collection for ACTION in the NTR was 42%. Urine and/or buccal-cell DNA samples (including 107 additional samples for epigenetic testing) are available for 1,495 twins (747 complete twin pairs), for whom 1,382 urine samples and 1,488 buccal-cell DNA samples were available. Twins have a mean age of 9.6 (SD age = 1.8) 47.5% of the participants were females (N = 710) and 82.6% belonged to a MZ twin pair (N = 1,234).
Objective 3.2: Unravel the genetic longitudinal architecture of aggression and estimate heritability as a function of age and gender
In order to fulfill objective 3.2 we took several approaches and used data of multiple ACTION cohorts. The main findings of the work under this objective are:
Genetic effects of childhood aggression are mostly stable, meaning that genes affecting childhood aggression are largely the same throughout childhood.
The heritability of aggression is similar in boys and girls, namely about 50-55% of the variability of the phenotype.
The indicators of aggression (i.e. questionnaire items) change in relevance throughout childhood (e.g. the item “threatens others” gains in importance). There are differences between boys and girls with respect to which items are the most relevant indicators of aggression. There is substantial differential heritability for the different indicators of aggression. For example, destruction of properties shows a very high genetic component during early childhood and fighting behaviors are more heritable in early adolescence.
A change in rater (mother vs. self-ratings) introduced complexity in the phenotype.
Measurement non-invariance across age can bias growth curve mean and variance estimates, and our quantification of this bias permits researchers to weigh the costs of using a sum score in longitudinal studies. Simulation results indicated that the genetic variance decomposition of growth factors is, however, not biased due to measurement non-invariance across age, provided the phenotype is measurement invariant across birth-order and zygosity in twins.
Across different informants, general and direct aggression at ages 12 and 14 predicted Antisocial personality disorder (ASPD) in a population-based sample.
Objective 3.3: Resolve patterns of comorbidity of aggression and emotional and behavioral problems as a function of shared genes and environment
In order to fulfill objective 3.3 we took several approaches and used data of multiple ACTION cohorts. The main findings of the work under this objective are:
Aggression co-occurs with the majority of other behavioral and social problems, from both externalizing and internalizing domains. We found neither rater differences nor differences across assessment instruments in co-occurrence patterns.
There were large similarities in co-occurrence patterns across European countries. Co-occurrence was also similar across age and sex, and if any change was observed, it indicated stronger correlations when children grew older.
For teacher ratings, correlations of aggressive behavior were high with other externalizing behaviors, and lower with internalizing problems. A negative association was seen with prosocial behavior.
Aggressive behavior at baseline was associated with higher BMI at follow-up. There was no evidence for reversed association. Aggressive behavior was prospectively associated with higher fat mass but not fat-free mass.
Over and above this list of findings of the ACTION project so far, there is also work in progress, including a study on the link between Insomnia and Aggression. Previous literature has shown a link between insomnia and all kinds of psychopathologies. We are conducting a study to test the hypothesis that a higher genetic risk for insomnia predicts aggressive behavior. With the latest summary statistics on an insomnia GWAS meta-analysis just released the analysists at each institution have just performed their prediction models and the meta-analysis is in progress.
Objective 3.4: Detect genomic regions of interest for aggression in children through genome-wide association (GWA) studies.
We took several approaches and used data of multiple ACTION cohorts and beyond. The main findings of the work under this objective are:
Overlap in item content across AGG measures ranged from absent (i.e. mutually exclusive) to moderate. Clinical concordance for the different AGG measures was very weak to weak. Rank correlations among continuous scores of AGG measures were moderate, suggesting stronger agreement when considering continuous scores compared to clinical cut-offs. Polychoric correlations revealed moderate to strong agreement between the different AGG measures. However, genetic correlations indicated substantial overlap in underlying genetic liability among the AGG measures.
A factor score meta-analysis can provide a gain in power over using sum score scores that were not directly comparable, provided the bi-factor model was adequately specified.
Our large meta-analysis of genetic associations with AGG from 29 cohorts was completed. It included 329,761 observations from 76,441 We report several suggestive loci on chromosomes 2, 3 and 13. The SNP-Heritability (h_SNP^2) of AGGoverall was estimated at 3,5%.
Rater-specific GWAMAs returned no genome-wide significant hits. Each rater-specific GWAMA, however, returned multiple suggestive loci. While some loci appear to be shared across multiple raters, none of the loci are common to all four rater-specific GWAMAs. Additionally, some loci are unique to their respective rater-specific GWAMA. The SNP-heritabilities range between 4% and 9.7% for different raters. Genetic correlations were moderate between AGGmother and AGGself (r_g=0.68; SE=0.10) and high between AGGmother and AGGteacher (r_g=0.83; SE=0.11). Interestingly, the genetic correlation between AGGself and AGGteacher was substantially lower (r_g=0.39; SE=0.12). The three rater-specific GWAMAs show largely the same genetic correlations with all outcomes.
Genetic correlations of childhood aggression with other traits give a meaningful picture of the genetic liabilities that are shared with childhood aggression: higher levels of AGG in children are associated with a much increased genetic risk of ADHD, and of developing a wide range of psychopathologies in adulthood. Children with high AGG have larger risks of loneliness, smoking, poor self-rated health, obesity and sleep problems.
Objective 3.5: Identify differentially methylated regions on the human genome related to differences in aggression using a genome-wide approach.
DNA for this objective was collected (see 3.1) in the same group of twin pairs and clinical cases as for the metabolomics projects. The samples have been measured on the Illumina EPIC 850K array. QC of the data has been performed with the pipeline developed by us in collaboration with the Dutch BBMRI-BIOS consortium (https://www.bbmri.nl/acquisition-use-analyze/bios/). Pipeline development and adjustment was required because Illumina stopped the production of the 450K array on the launch of their new Epic (850K) during the ACTION project. To adapt data processing and analysis pipelines for the EPIC array, we performed a study on DNA samples (from existing NTR DNA collections for twins with buccal swaps and aggression phenotyping) measured on the Illumina EPIC array. The samples were from 53 monozygotic twin pairs (age 0-10), scoring high or low on aggressive / rule breaking behavior. For one twin pair, the same DNA sample was run twice on to examine test-retest reliability of the array. We performed the first genome-wide analysis of monozygotic (MZ) twin correlations and mQTLs on data obtained with the Illumina MethylationEPIC BeadChip (EPIC array) and compared its performance to the Illumina HumanMethylation450 BeadChip (HM450 array) for buccal-derived DNA. We conclude that the contribution of familial factors to individual differences in DNA methylation and the effect of mQTLs are larger for novel EPIC probes, especially those within regulatory elements connected to active regions specific to the investigated tissue.
Epigenome-wide association study for Aggression: While the new DNA samples were being collected and measured, we performed the first large EWAS meta-analysis (EWAMA) on existing data of aggressive behavior and genome-wide DNA methylation assessed with the Illumina 450k array (mainly) or the EPIC array. We contacted all possible collaborators around the world for the availability of Methylation and aggression data, which resulted in the inclusion of 21 cohorts with data for 14,434 peripheral blood samples (age at blood sampling 7 to 70) and 2425 cord blood samples in the EWAMA. For follow-up analysis we collaborated with the other consortia that received funding in the FP7 Work program HEALTH.2013.2.2.1-3: Paediatric conduct disorders characterised by aggressive traits and/or social impairment: from preclinical research to treatment. Data in two clinical cohorts were available in NeuroIMAGE and FemNAT-CD and brain methylation data from rodent models of aggression from the MATRIX consortium. For the discovery analyses, the age at phenotypic assessment ranges from 4 to 68 and multiple instruments were used to assess aggressive behavior or conduct problems.
EWAMAs were performed for whole blood, cord blood, and for all data combined. These analyses identified 48 top CpGs for which we performed enrichment analyses. The top ten most strongly enriched traits indicate that CpGs associated with aggressive behavior overlap with CpGs previously associated with smoking, maternal smoking, and other chemical exposures (e.g. perinatal exposure to polychlorinated biphenyls (PCBs) and polychlorinated dibenzofurans (PCDFs)). Further overlap includes CpGs associated with alcohol use, cognitive function, educational attainment, ageing, and metabolic traits. Next steps involve integration of the EWAMA findings with the novel EPIC methylation data collected in D3.1. The genome-wide summary statistics from this project will be made available to the scientific community.
Epigenome-wide association study for ADHD symptoms: On existing data we also performed the first epigenome-wide association study meta-analysis of adult attention-deficit/hyperactivity disorder symptoms, based on peripheral blood DNA methylation (Infinium HumanMethylation450K array) in three population-based adult cohorts: the Netherlands Twin Register (N = 2258, mean age 37 years), Dunedin Multidisciplinary Health and Development Study (N = 800, age 38 years), and Environmental Risk Longitudinal Twin Study (N = 1631, age 18 years).
One epigenome-wide significant differentially methylated position was detected in the Dunedin study, but meta-analysis did not detect differentially methylated positions that were robustly associated across cohorts. In region-based analyses, six significant differentially methylation regions (DMRs) were identified in the Netherlands Twin Register, 19 in the Dunedin study, and none in the Environmental Risk Longitudinal Twin Study. Of these DMRs, 92% were associated with methylation quantitative trait loci, and 68% showed moderate to large blood-brain correlations for DNA methylation levels. DMRs included six non-overlapping DMRs (three in the Netherlands Twin Register, three in the Dunedin study) in the major histocompatibility complex, which were associated with expression of genes in the major histocompatibility complex, including C4A and C4B, previously implicated in schizophrenia.
Results for Objectives from our work package Gene-Environment interplay
Objective 4.1: Identify and replicate ‘pure’ environmental risks that have their effects on the development of aggression independent of genetics
ACTION twin analyses have contributed to the research literature demonstrating that most of the environmental variance in aggression and other behavior problems is not shared by children growing up in the same family. We have exploited the powerful identical (MZ) twin differences design to identify pure environmental factors contributing to aggression. We have now completed 10 extensive interviews of MZ twin pairs who are discordant at 1.25 standard deviations for measures of aggression assessed at age 16. Interviewers explored the family, school and friend-wide factors that each member of a twin pair separately attributed to their discordant aggression scores. Crucially, each member of a twin pair reported on their own aggressive behavior but also on their co-twin’s aggressive behavior, which allowed for corroboration of environmental contributors across raters. Analyses of our in-depth interview data suggest adolescents’ aggressive behaviour is likely to be influenced by several risk factors spanning multiple areas of their lives. As a result, we harnessed the predictive power of multiple risk factors, namely: environmental niches (school, home and neighbourhood characteristics); (2) growth and health (BMI, pubertal development, physical and mental health); and (3) psychological resilience (psychological adjustment, wellbeing and motivation) to create a multidimensional risk score (MRS) for 5,635 twin pair members of the Twins Early Development Study (TEDS). We conducted twin model-fitting analyses of these data.
MRS predicted a similar portion of variance in concurrent aggression at both ages 12 and 16 (explained variance R2 = .10 and R2 = .11, respectively). The aetiology of the association between MRS and aggression differed between early and late adolescence. While shared environment explained most of the covariance at 12 (51%), genetic factors were the main source of covariance at 16 (50%). Although non-shared environment was the main factor in the aetiology of aggression at 16 (e2 = 72%), it was only minimally implicated in the overlap between MRS and aggression. Our interviews with 10 MZ pairs highly discordant for aggression suggest that non-shared environments are largely idiosyncratic, such as differences in romantic relationships, differences in the use of oral contraceptive or differential responses to parental separation. Thus, risk factors of aggression span multiple domains of adolescents’ lives. Although MRS predicted a significant portion of variance, the effect is modest, emphasising the need for future research to identify additional risk factors. The observed lack of non-shared environmental overlap with aggression suggests that non-shared environment is idiosyncratic. This in-depth analysis of the individual and collective role of socio-ecological, physical and psychological characteristics to variation in aggression is likely to provide essential new knowledge on key targets for environmental interventions aimed at reducing the personal and societal burden associated with aggressive behaviour.
Objective 4.2: Identify and replicate environmental effects on the development of aggression that depend on genetic propensities (GxE)
GxE using Twin Method: In order to address GxE we first we examined phenotypic correlations between environmental measures separately at ages 7, 9, 12 and 16 and verbal and physical aggression at age 21. We then compacted the analyses in a multiple regression approach that examined the combined prediction from all childhood environmental measures at each childhood age on verbal and physical aggression at age 21. At each age and across ages, we created a novel ‘poly-environmental’ risk score for age-21 aggression. To test the joint prediction of childhood environmental measures on age-21 aggression, we used a leave-one-out multiple regression approach. That is, regression weights from 90% of the sample were used to create a poly-environmental risk score for the other 10% so that scores were derived from an independent sample. This poly-environmental score was calculated at ages 7, 9, 12 and 16. We then explored whether the heritability of aggression differs at different levels of the poly-environmental risk score using a continuous GxE analysis (Purcell, 2002). For example, does high poly-environmental risk for aggression lower heritability by overwhelming genetic propensities?
Results indicated that Poly-environmental scores at each age correlated modestly with aggression at age 21 (average correlation = 0.2). The correlations were of similar magnitude for physical and verbal aggression. These are ‘main effects’ of the childhood environment on adult aggression.
We examined the extent to which the poly-environmental risk scores moderated the manifestation of genetic and environmental effects on aggressive behaviour during childhood and adolescence. The results showed significant GxE interaction effects during childhood. At age 7, the twin heritability of aggression was observed to increase at higher levels of exposure to poly-environmental risks factors, going from ~.20 at lower levels of environmental risk (-2SD) to ~.60 at high levels of exposure to environmental risk (+2 SD). A similar pattern of results was observed at age 9, where the heritability of aggression was ~.30 at low levels of exposure to environmental risk and ~.60 in twins exposed to higher levels of poly-environmental risk. However, these interaction effects did not replicate in adolescence, when the twins were 12 and 16 years old. At these ages, twin heritability estimates remained consistent (between .3 and .4) at different levels of exposure to poly-environmental risk. We conclude that childhood environments predict adult aggression but the evidence for moderation effects of genes on adult aggression is limited. We detected significant GxE for aggression at ages 7 and 9, but this did not replicate at ages 12 and 16 in the TEDS sample. Further replication in other ACTION samples is planned to test for the specificity of detected GxE to early childhood years.
GxE using Polygenic Score method: Polygenic scores derived from major GWA studies of adult psychiatric disorders were used to create a polygenic score for ‘p’, a general factor of psychopathology, for each individual in TEDS. We then assessed GxE interaction between the polygenic score for p and the environmental measures in terms of their prediction of aggression. Unlike twin analyses, we were able to investigate GxE for all of our environmental measures regardless of whether they were child-specific or family-wide. We tested poly-environmental scores at ages 7, 9, 12 and 16 (see Study 1) in our analyses of GxE. We did not find significant GxE effects. That is, the prediction of aggression from the environmental measures was not conditional on the polygenic score for p. Thus, for the measures used in our analyses, we can conclude that GxE for aggression does not emerge in systematic analyses using polygenic scores as well as using the twin method. However, novel environmental factors, most notably therapeutic interventions, could show stronger interactions with genetics. The key advance needed for polygenic score research on aggression is to create more powerful polygenic scores, but this advance depends on increasing the SNP heritability of aggression, which is low (<10%).
GxE using novel methods: Because of difficulties in identifying GxE, we developed new approaches using atheoretical machine-learning approaches to detect interactions across diverse environmental measures, polygenic scores, and phenotypic measures. First, we perform Elastic Net regularization, including covariates such as age, sex and 10 genetic principal components, independent variables (polygenic scores and environmental predictors), and the bivariate interaction terms for covariate*predictor and predictor*predictor. Second, we fit a random forest regression to the data, including covariates and predictors (PGS and measured environments), leveraging the ability of regression trees to automatically detect interaction patterns in the data, as well as non-linear effects. For each phenotype, all models are trained in 80% of the data and tested for performance (with respect to prediction accuracy) in the remaining 20%. Analyses include proximal and distal environmental measures such as life events, home environment, and socio-economic status, along with polygenic scores for major psychiatric disorders (e.g. schizophrenia), cognitive traits (e.g. intelligence) and anthropometric traits (e.g. body mass index). We conducted these analyses separately for each trait. Prediction models were constructed for behaviour problems (including aggression), psychiatric disorders and anthropometric domains. For every phenotypic outcome, we fit linear and decision-tree based models integrating environmental and polygenic predictors, assessing their predictive accuracy while testing for (multivariate) interactions. Preliminary results from our prediction models indicate that this new approach yields many suggestive GxE but of very small effect size. When the ACTION+ GWAMA results are available, we will include a polygenic score derived from the summary statistics in these analyses. Follow up linear models (such as OLS regression) for all detected interactions are currently being conducted in replication cohorts in ACTION.
Objective 4.3: Identify and replicate genetically driven experiences associated with the development of aggression in which children select, modify or create environments correlated with their genetic propensities towards aggression (rGE)
rGE using Twin Method: We systematically explored gene-environment correlation (rGE) using bivariate twin analyses for a wide range of environmental measures as they predict self-reported verbal and physical aggression at age 21. We found that 90% of the correlation between the poly-environmental risk score and aggression could be accounted for by genetic factors. Similar results were found in bivariate twin analyses of those individual childhood environmental measures that predicted adult aggression. These findings strongly implicate rGE in the prediction of adult aggression from childhood environments.
rGE using Polygenic Score Method: Twin analyses of rGE are limited to environmental measures that differ within families (i.e. child-specific measures). Here we used a polygenic score for ‘p’ and a general factor of psychopathology in analyses of rGE. We first asked if the polygenic score for ‘p’ was, as expected, associated with our measure of aggression at age 21. We then tested for an association between the polygenic score for ‘p’ and the poly-environmental risk score for aggression. We found that the polygenic score for ‘p’ was significantly associated with aggression but predicted less than 0.5% of the variance in aggression, which is comparable to the variance explained for other measures of behaviour problems. A correlation of .05 was found for the association between the polygenic score for ‘p’ and the poly-environmental risk score for aggression, indicating rGE. This polygenic score for ‘p’ correlates as highly with the poly-environmental risk score for aggression as it correlates with measures of behaviour problems including aggression. A major limitation of these analyses is that the polygenic score is only weakly associated with aggression, which makes it difficult statistically to show that the association between childhood environmental risk and adult aggression. The key advance needed for polygenic score research on aggression is to create more powerful polygenic scores, but this advance depends on increasing the SNP heritability of aggression, which is low (<10%). Nonetheless, for the measures used in our analyses, we can conclude that rGE for aggression is widespread in systematic analyses using polygenic scores as well as using the twin method.
rGE using DZ twins: Polygenic score prediction among unrelated individuals may include contributions from both direct genetic effects and also indirect effects due to passive GE, as well as population stratification and assortative mating. Here we used the within-family association between sibling differences in polygenic scores and sibling trait differences, which is free of such indirect effects of shared environment, to test for rGE. The most relevant finding is that evidence was found for passive rGE for cognitive traits (intelligence and educational achievement) but not for any other traits such as personality or behaviour problems, including physical and verbal aggression. The implication for aggression is that passive rGE is unlikely to be an important factor in the origins of individual differences in aggression.
Objective 4.4: Apply GCTA methods, GWA genotype data, and polygenic predictors of aggression phenotypes to confirm the results of twin-family analysis of GxE and rGE and to identify some of the genes responsible for GxE and rGE
By leveraging a SNP-derived genetic relationship matrix (GRM) we compared genetic similarities between unrelated individuals to similarities on verbal and physical aggression to infer the contributions of SNPs to the variation in these traits. SNP heritability analyses for the BAQ self-report measure at age 21 yielded an exciting result: the SNP heritability for verbal aggression was 32%, much higher than typical SNP heritability estimates of about 10%. However, when we tested this result in the Queensland sample, which included the same BAQ measure, the result did not replicate, demonstrating the value of ACTION for immediate replication research. SNP heritability for aggression is about 10%, typical of other behaviour problems, which is important because SNP heritability is the ceiling for GWA studies.
Objective 4.5: Extend analysis of epigenetics (WP3) to investigate GxE and rGE
DNA methylation data from the ACTION EWA meta-analysis (EWAMA) of aggression were used to create polygenic scores for TEDS twins by weighting the corresponding top loci that emerged from the EWAMA (i.e. SNP-CpG associations) by their effect sizes (z-scores), using the –score function in Plink (epitot). We also created two additional scores, one based on the CpGs positively associated with aggression (epipos) and one with the ones negatively associated with aggression in the EWAS (epineg). Our SNP polygenic score for the top methylation sites was not significantly associated with verbal, physical or total aggression at age 21. This polygenic score was also not correlated with our poly-environmental measure (rGE), nor did it mediate the modest correlation between the poly-environmental measure and aggression. The EpiGPS scores were not associated with aggression, aggression subtypes, polygenic p, or our multi-environmental scores. Ongoing analyses are investigating GxE and rGE using epigenetic scores from the ACTION EWA of aggression.
Results for Objectives from our Biomarker work package.
Objective 5.1: to identify biomarker profiles to diagnose aggression
A review of the literature was performed to identify known biochemical biomarkers for human aggressive behavior . This identified four classes of biochemical biomarkers potentially involved in human aggressive behavior: 1) inflammation markers (e.g. interleukin 6, C-Reactive Protein); 2) neurotransmitters (e.g. 5-HIAA, HVA, glutamate); 3) lipids (e.g. HDL, LDL); and 4) hormones (e.g. cortisol, testosterone). Methylation markers have also been suggested to be associated with aggressive behavior. Furthermore, Dekkers et al. (Genome Biol., 2016) showed that methylation and blood lipid levels are not independent and that differential methylation can be a consequence of variation in blood lipid levels. Therefore, we hypothesized that the methylation levels of such loci in blood can inform us if aggression is associated with long-term exposure to lipid level. If this is the case, we expected to find that loci where methylation levels were influenced by lipid levels show differential methylation in aggressive individuals. Such loci might complement classic lipid level measures as a biomarker for lipid-related disturbances in aggression.
As a first step, we examined the association of lipid levels and related biomarkers with aggression in existing data from a large adult population cohort (N = 5,588), including 31 monozygotic (MZ) twin pairs who were discordant for aggression, as well as 12 extremely discordant MZ pairs. Biomarkers were not significantly associated with aggression in the population cohort. In the discordant MZ pairs we identified significant within-pair differences for glucose and marginally significant differences for lipids and cytokines, with the more aggressive twin showing lower levels of glucose and LDL and higher levels of fibrinogen, C-reactive protein and interleukin-6. The analysis of epigenetic data in the MZ pairs discordant for aggression did not show enrichment for lipid cytosine guanine dinucleotides (CpGs) and we observed no enrichment of lipid CpGs in an epigenome-wide association study of aggression in the population cohort.
In the Biochemical Study we analyzed urine samples of 222 children (mean age = 9.9; SD age = 1.6) selected for the absence or presence of aggression problems. We reclassified the case-control status of twins based on their mother-rated aggression scores at the time of urine and DNA collection. This classification identified 110 cases, 95 controls (17 children without new status). To identify potential biomarkers for childhood aggression we developed LASSO GLM models for the organic acids and amines metabolomics platforms. These models identified no biomarkers based on the amines platform and selected 4 metabolites (α-hydroxybutyrate, succinic acid, aspartic acid and uracil) based on the organic acids’ platform. The predictive accuracy (AUC) of the organic acids model is relatively was 67% and including these metabolites in a model with sex and age did improve the predictive accuracy compared to a model including only sex and age. Univariate follow-up analysis including age, vitamin use and batch effect as covariates found a significant association with aggression for α-hydroxybutyric acid (β = 0.31 p = 0.04).
In the Main Study we analyzed urine samples from 1,348 twins (mean age = 9.6; SD age = 1.8) selected for their concordance or discordance for aggression problems. In addition, the Main Study included 186 children of 6 to 13 years of age (mean age = 9.6; SD age = 1.8) referred to the clinic. To identify biomarkers for childhood aggression in the Main Study we used a three-step approach: 1) between-family analyses; 2) within-family analyses and 3) replication analyses.
For the median-normalized metabolites, prior to correction for multiple-testing, the between-family analyses showed significant negative associations of L-methionine-sulfoxide (estimate = -0.003; Robust SE = 0.002; p = 0.03) and glyceric acid (estimate = -0.001; Robust SE = 0.001; p = 0.04) with childhood aggression. We observed no significant differences in metabolite concentrations between aggressive and non-aggressive twins for the top 25% of the amines (16 metabolites) and organic acids (5 metabolites) in the within-family analyses. Congruent directions of effect in the between- and within-family analyses were observed for 2 of the 5 top 25% organic acids and 7 of the 16 top 25% amines. Repeating the within-family analyses with only discordant MZ twins improves the congruency of the direction of effect among the analyses to 16 of the 21 top 25% metabolites. Replication of the top 5 metabolites across both platforms (3-methyladipic acid, cystathionine, cysteine, glyceric acid and homocysteine), in 186 clinical cases and 93 twin pairs concordant low for childhood aggression, confirmed that amines and organic acids were not significantly associated with childhood aggression in the ACTION project.
For the density-normalized metabolites and biomarkers, after correction for multiple testing the between-family analyses showed associations of childhood aggression with O-phosphoserine (estimate = 0.33; Robust SE = 0.09; p = 0.02) Gamma-L-glutamyl-L-alanine (estimate = 0.31; Robust SE = 0.09; p = 0.02) and Gamma-Glutamylglutamine (estimate = -0.16; Robust SE = 0.05; p = 0.03). The concentrations for these 3 metabolites were not significantly different between aggressive and non-aggressive twins. Replication analyses were performed for the top 5 metabolites most associated metabolites with consistent direction of effect in the between- and within-family analyses in 186 clinical cases and 93 twin pairs concordant low for childhood aggression. We observed that norepinephrine (mean cases = 0.37; mean controls = -0.15; p = 3.03x10-5) saccharopine (mean cases = 0.20; mean controls = -0.08; p = 0.01) ethanolamine (mean cases = 0.21; mean controls = -0.08; p = 0.01) and gamma-L-glutamyl-L-alanine (mean cases = -0.01; mean controls = -0.23; p = 0.03) different significantly among cases and controls after correction for multiple testing, with congruent direction of effect across all 3 analyses.
In conclusion, based on the existing literature we identified 4 classes of potential biomarkers for (childhood) aggression: 1) lipids, 2) inflammation markers, 3) neurotransmitters and 4) hormones. With the work performed in ACTION WP5 we find further suggestive evidence for the role of inflammation markers, lipids and amino acids involved in neurotransmitter pathways in aggression. In children we identified metabolites of the cysteine metabolism pathway, involved in oxidative stress, as among the most promising aggression biomarkers and in adults we found cytokines to be marginally significantly related to aggression differences in extremely discordant MZ pairs. LDL cholesterol levels were lower (marginally significantly) for the MZ twins scoring high on aggression as compared to their co-twin scoring low on aggression. Finally, the top 25% most associated metabolites in ACTION’s Main Study included multiple (essential) amino acids which are involved in neurotransmitter pathways, or may act as neurotransmitters themselves, such as the metabolites included in the cysteine metabolism pathway.
Objective 5.2: to identify biomarkers to differentiate between sub-groups of aggression
The most common subtypes of aggression studied to date are verbal vs. physical aggression, impulsive vs. premeditated aggression or reactive vs. proactive aggression. Our literature review suggested that lipids, neurotransmitters and hormones are associated with subtypes of aggression, however, the literature does not suggest any clear pattern of biochemical biomarkers associated with subtypes of aggression. For example, decreased cortisol levels have been associated with both reactive and proactive aggression, while increased cortisol levels have also been associated with reactive aggression.
In order to add the additional steroid targets suggested by literature to the target list measured in the biochemical study, development on a novel steroid platform measuring both conjugated and unconjugated forms was undertaken. Steroids in this platform have a wide range of polarities and extensive development work was necessary. Getting the proper standards for the targeted LC-MS/MS was difficult as many steroids are controlled compounds and took a longer time than originally expected, but the method is now being validated and the samples are measured. The target list that has been finally chosen is based on input from various discussions within ACTION, with the scientific advisors of ACTION, and other experts in the field, and the target list is based on availability of the steroids as reference compound. The developed platform will now be used for samples collected as part of the ACTION project and will also be available for future studies.
To identify subgroups for aggression we defined subgroups based on our work of the comorbidly with other psychiatric traits, specifically symptoms of attention-deficit/hyperactivity disorder (ADHD). In 1,329 twins we investigated the role of biomarkers and metabolites on aggression comorbid with ADHD-symptoms. This comorbidity phenotype was classified as follows: “low aggression and low ADHD-symptoms”, 2) “high aggression and high ADHD-symptoms”, 3) “low aggression and high ADHD-symptoms”, and 4) “high aggression and low ADHD-symptoms”. GEE analysis of the aggression comorbidity phenotype showed that 14 median-normalized metabolites levels associated with aggression and/or ADHD-symptoms as compared with controls (children with low aggression and ADHD-symptoms). Higher levels of 3-methylhistidine, adipic acid, anserine, glyceric acid and histamine, as well as lower levels of cystathionine and cysteine were observed for children with high aggression and high ADHD-symptoms. Lower levels of 3-hydroxybutyric acid, 3-methyladipic acid and fumaric acid were observed for children with high aggression but low ADHD-symptoms. Low aggression and high ADHD-symptoms were associated with increased levels of L-isoleucine and L-threonine. Results for L-leucine and L-methionine-sulfoxide were inconclusive as higher metabolite levels were observed for multiple subgroups.
GEE (generalized estimating equation) analyses on the density-normalized metabolites and biomarkers indicated that lower taurine, gamma-glutamylglutamine, epinephrine and 3-hydroxyisobutyric acid levels and higher o-phosphoserine levels were significantly associated with high aggression and low ADHD-symptoms. Lower levels of citric acid and DL-3-aminoisobutyric acid were observed for children with low aggression and high ADHD symptoms. However, results were inconclusive for cysteine, fumaric acid, glycylglycine, homocysteine, malic acid, glutathione and cystathionine as these were associated with multiple subgroups.
In conclusion, with the work performed in ACTION WP5 we find further suggestive evidence for the role of amino acids involved in neurotransmitter pathways in aggression comorbidity. We found that levels of the essential amino acids L-isoleucine and L-threonine were associated with ADHD-symptoms, but not with aggression and observed that several metabolites, including metabolites of the cysteine metabolism pathway, were associated high aggression and ADHD-symptoms. Furthermore, the suggestive role of the cysteine metabolism pathway suggests that inflammation (oxidative stress) is also involved in aggression and ADHD-symptoms.
Objective 5.3: identify treatment options for (different sub-groups) of aggression
The amino acids isoleucine, leucine, methionine, tyrosine and valine had lower levels in children with high aggression as compared with low aggression (Deliverable 5.1). All of these amino acids are (conditionally) essential. Supplementation of these (conditionally) essential amino acids to increase their levels might help to reduce childhood aggression. Specifically, supplementation of methionine may be beneficial to alleviate aggression in children. Methionine is a precursor in the cysteine metabolism pathway. Methionine supplementation therefore might increase homocysteine levels which in turn might rebalance the disturbed cysteine metabolism pathway. With the balance in the cysteine metabolism pathway restored the potential disrupted antioxidant glutathione levels may be restored. Therefore, methionine supplementation might restore the balance between antioxidant and reactive oxygen species thus alleviating oxidative stress. ‘Subtype’ analyses show that restoring the balance to the cysteine metabolism pathway may be especially beneficial for those children with comorbid high aggression and ADHD-symptoms.
Results for Objectives from our work package Novel treatment and prevention strategies.
Objective 6.1: Development of evidence-based model on effectiveness of existing preventative and treatment strategies for paediatric aggression through meta-analyses (1. Prevention, 2. Intervention)
To extend existing knowledge on prevention and intervention of childhood aggression, we did a literature synthesis of 72 systematic reviews and meta-analyses on treatment effectiveness for childhood aggression. For universal and selective prevention, treatment effectiveness was absent or weak; for indicated prevention and intervention, treatment effectiveness was mostly small or medium. Furthermore, most moderators of treatment effectiveness had no effect in the majority of studies (i.e. child age, child gender, implementation to individuals or groups, different treatment programs, person implementing the treatment, and session related factors or treatment intensity) or mixed effects (i.e. socioeconomic status, type of treatment, informant, research quality). The only two significant moderators comprised of positive effects of higher pre-treatment levels of aggression and parental involvement. Findings suggested that future research distinguishing between targets of the treatment (i.e. factors associated with childhood aggression; universal and selective prevention vs. present aggressive behaviors; indicated prevention and intervention) would be promising. In addition, more knowledge on parental influences is needed, including parental psychopathology, parenting style, and heritability .
Objective 6.2: Integration of insights from WP2-WP5 in a comprehensive theoretical framework to inform novel prevention and intervention efforts .
We collaborated on a project that combined data from CATSS and NTR (total sample size was 62,227 children) to predict childhood aggression using a novel methodological approach. The outcome measure was the phenotypically harmonized overt/physical aggression factor score created within Action. The large set of predictor variables encompassed demographics, prenatal characteristics, physical development, parental education level, life events, and mother-reported behavioral symptoms. To avoid capitalization of chance, data were partitioned in four parts for the different analysis steps. These included 1) exploratory data analysis and tuning meta-parameters for the data mining, 2) fitting increasingly complex data mining models to assess which predictors had which types of effects (i.e. linear, nonlinear, interaction), 3) assessment of model performance and importance of the predictor variables, and 4) fitting a confirmatory prediction model of aggression that integrated the results of the data mining analyses. The resulting multi-group model accounted for interactions with sex and cohort and confirmed linear main effects of variables measuring behavioral symptoms (e.g. related to non-physical aggression, attention-deficit/hyperactivity disorder, and conduct disorder), maternal smoking during pregnancy, parenting, and proportion of life events. The most important predictors comprised behavioral symptoms such as arguing, being easily distracted, and hyperactivity. Findings were in line with previous research, yet weaker, likely due to the simultaneous analyses of many predictors. These easily observable predictive behaviors may act as targets for early detection and prevention of childhood aggres. These findings confirmed that childhood aggression cannot be viewed in isolation from other emotional and behavioral problems.
Objective 6.3: Development of risk assessment charts together with guidelines to improve clinical decision making
Development of childhood aggression is shaped by interactions between individuals and multiple contexts; a non-exhaustive overview can be found in the Table below. These contexts range from the proximal (e.g. family) to the more distal contexts (e.g. policies for child and adolescent mental health). To examine the extent to which national-level policies for child and adolescent mental health (CAMH) are associated with individual-level aggressive behavior, we analyzed data from 172,829 adolescents (aged 11 to 15 years) in 30 European countries from the Health Behavior in School-aged Children study. Information on national-level policies for CAMH was gathered from renowned statistical institutes and included availability of epidemiological data, the number of CAMH facilities, investment in family benefits, and investment in education. Multi-level analyses revealed lower adolescent aggressive behavior in countries with more CAMH policies. This association held when taking into account other national-level indicators, including adult violence, adult well-being, and income inequality. We translated these findings and previous literature into guidelines to improve clinical decision making.
An established risk factor for childhood aggression is lower socioeconomic status (SES). There is, furthermore, some evidence that SES has a moderating effect on the causes of individual differences in childhood aggression, with lower heritability estimates in low SES groups and a stronger contribution of environmental factors. In collaboration with researchers from WP3 and WP4, we conducted a study on the moderating effect of socioeconomic status (SES) on the genetic architecture of childhood aggression among 7-year-old twins in the Netherlands and the United Kingdom. We found that, compared to high SES, for low and medium SES the standardized contribution of the shared environment to individual differences in childhood aggression was higher. This finding suggests that aggressive children from low and medium SES may benefit more from interventions targeting their surroundings than aggressive children from high SES.
Our review on the effectiveness of interventions highlighted the importance of parent and family factors. Family factors e.g. inter-parental violence, family conflict and harsh parenting, contribute to childhood aggression by diminishing children’s self-control. At the same time, children with low self-control evoke more harsh and negative parenting behavior and family conflict. In addition, following up on our review, we contributed to a study (WP2) that investigated if parental mental disorders (MD) increase risk of future maladjustment among children who early-onset disruptive behavior (DB). In a large, longitudinal study, parents of 9-year-old children reported on child DB, while a patient registry was used to determine parental MD. At follow-ups at 15 (N = 6319) and 18 years (N = 3068) information about various problems was collected via registries, parent- and, self-reports. Results revealed that parental MD puts 9-year-olds with DB at risk for negative outcomes in adolescence. Additionally, paternal MD was a better predictor than maternal MD, regardless of child DB at age 9, suggesting that fathers should be given increased attention in future research. The findings also suggest that the effectiveness of prevention and treatment efforts for children with DB can be enhanced with additional screening and, if necessary, treatment of parental MD.
Objective 6.4: Assessment of the empirical theoretical framework and risk assessment charts with respect to ethical, legal and social implications (ELSI)
To address the ethical, legal, and social implications (ELSI) of the research conducted within ACTION, we contacted all partners and reviewed all publications and ACTION manuscripts. The outcomes are discussed in Deliverable 6.3. In brief, ethical implications of ACTION comprised of the concept of childhood aggression (i.e. multidimensional disorder, highly co-morbid with other emotional and behavioural problems likely due to a shared genetic liability, greater agreement among assessments of aggression using a continuous approach), patients’ understanding (i.e. when genetics and biomarkers are implemented in treatment all stakeholders need to understand what it means), and the potential for stigmatization and the harmful effects of labelling (i.e. prediction models need to reach high precision and accuracy before the benefits outweigh costs of misclassification). Legal implications concerned redefinition of existing regulatory policies (i.e. more research is needed to understand the association between policies for child and adolescent mental health and aggression), professional ethical guidelines (i.e. academic experts expressed a need for national-level and Europe-level guidelines for severe behaviour problems, with reinforcement for implementation, tailored to daily practice), and inclusion of biomarkers in clinical decision making (note that biomarkers with such predictive value have not yet been identified in ACTION research). Social implications comprised stigmatization and labelling effects (i.e. subpopulation identified for heightened risk of aggression is children whose parents have a psychiatric disorder, also an important factor for treatment effectiveness), and disparity, diversity, and equity (i.e. results may not generalize to minority groups, etiology of aggression differs among SES strata so may also differ for minority groups). Although the research within ACTION does not yet have implications that can directly be translated into ELSI, the built collaborations and the wealth of data do provide sufficient tools to work towards improvements in prevention and intervention efforts and identification of actionable ELSI in the future.
RISK FACTORS ASSOCIATED WITH AGGRESSION IN CHILDHOOD AND ADOLESCENCE
Individual factors
Biological characteristics Psychological and behavioral characteristics Demographics
Biomarkers including metabolomic profiles
Genetics (heritability, GxE)
Genetics (GWAS, epigenetics)
Complications/risks during pregnancy (e.g. maternal smoking; birth weight; maternal age)
Resting heart rate
Puberty
Body size
Attention problems, lack of behavioral control (impulsiveness, low executive functioning).
Arguing, being easily distracted, hyperactivity
Low IQ
Low educational performance
Low school commitment
Drug use (esp. at an early age)
Callous Unemotional traits Sex
Age
Social factors
Family influences Parental conflict Peer influence
Parenting:
Parental warmth
Control
Harsh discipline, spanking
Monitoring, supervision
Neglect Divorce
Domestic violence Having delinquent friends or siblings
Drug use
Family characteristics
Demographics and structure Parental characteristics
Many children in the family
Single (teenage) mom
Stepfamilies
Parental age
Parental SES Parental psychopathology
Community factors
Neighbourhood SES
Neighbourhood violence
Urban vs. rural area
Social cohesion
Societal factors
ISocial factors Cultural factors
Income inequality
War
Natural disasters
Terrorist attacks Social cohesion
Norms or values (honor culture)
Violent media
Results for Objectives from our work package Dissemination
Objective 7.1: Definition of a recognizable and global project identity : GenerACTION.
A Project’s Logo: The project logo was designed and appointed in all the presentations and dissemination material produced and shown by the consortium. The consortium released banners, headers, standard “deliverable template” and a standard “presentation template” for uniform the communications documents and a standard “poster template”.
The project’s official website is available at: http://www.action-euproject.eu. It consists of a public area and a restricted section and contains data protocols, news, events, media, presentations, photo galleries, partners and staff’s information. The website has been constantly updated and extended during the project. The website’s traffic has been constantly monitored with the help of statistic tools. Overview of the visits during the whole project:
-More than 24000 page views and More than 5500 unique users
Sessions to the website are constantly growing from June 2015 to June 2019 (the website was released on December 2014, the monitoring started on June 2015).
-Geographical distribution of users.
Countries: 125 Countries reached. USA is the country with the highest number of users, meaning:
- the ACTION project is currently having a good impact also out of Europe (and also out of the countries with more effort in dissemination);
- research studies on aggression in children are particularly relevant in the USA.
-Cities: 1250 Cities reached. New York and Paris are the (known) cities with the highest number of users, despite no partners of the consortium are based in New York or in France.
Project presence in the social media channels.
The presence of the project on youtube has been ensured with a series of Interviews and videos.
15 videos from the ACTION final meeting (13 Interviews, 1 general video of the event, 1 video containing an award ceremony: best talk given by PhD candidates during the ACTION final event), May 6-8, Sardinia, Italy.
The interviews are available here: http://www.action-euproject.eu/content/videos-and-interviews-action-final-dissemination-meeting
4 Interviews to ACTION staff members and a total of 22 videos showing the ACTION logo. ACTION was a sponsor of the "13th International Workshop on Neonatology", Cagliari, Italy, October 25-28, 2017. A video gallery in the official Youtube page of the event is available and contains the following 4 videos focusing on ACTION, and other videos showing the ACTION Logo.
Interview with Prof. Dorret Boomsma - Main results of the ACTION project: https://youtu.be/6jmwszJXMLM
Interview with Prof. Fanos - The ACTION project's goals
https://youtu.be/PL5NiepwWto
Interview with Dr. Mirko Manchia - Risk factors for aggression
https://youtu.be/ewbbh-UoHU4
Interview with Prof. Luigi Atzori - Metabolomic approach and aggression, in Italian language: https://youtu.be/-9SOaTicUVM
3 Interviews to ACTION staff members and a total of 11 videos showing the ACTION logo. ACTION was a sponsor of the "14th International Workshop on Neonatology", Cagliari, Italy, October 24-27, 2018. A video gallery in the official Youtube page of the event is available and contains the following 3 videos focused on ACTION, and other videos showing the ACTION Logo.
The project’s presence on Twitter and Facebook has been ensured through posts and tweets sent by existing social profiles, using the hashtag #actionproject (according with the notes of the Ethical/Scientific Advisory Board “The Board requested however that Wkpg #7 limit the number of social media used (e.g. Facebook) due to the vulnerability of children and possible privacy violations”, the presence on the social media is limited to announcements and news).
Objective 7.2: Update participations and participating agencies InterACTION.
The consortium developed an online restricted area within the official project’s website: it contains contributions, opinions, shared documents and comments of the staff’s members.
The consortium created ten mailing lists for general@ action-euproject.eu (for all the partners, it can be used for general communications regarding the project (e.g. relevant milestones, events, publications, etc.) and for wp leaders, and for WPs
The consortium developed a Newsletter management system within the website and edited 5 Newsletter Issues (4 regular issues and a Special NTR Newsletter ISSUE from the beginning of the project to the present day). The newsletters were released and directly sent via e-mail to a total of 201 subscribers (including stakeholders, students, researchers):
All Newsletters issues are also available for download in the ACTION website (http://www.action-euproject.eu/content/action-newsletters). They have been also shared by the members of the consortium via private e-mail messages to researchers and stakeholders.
Training events for Junior staff and workshops organized by the consortium (lectures given within workshops and conferences managed by other organizations are not included here but mentioned in the related Work Packages or in the objective 7.4). Examples include:
The second ACTION consortium meeting hosted a Workshop titled "Beyond the Genetics of Aggression; From Clinic to Research and Back". It took place in the Old Meeting Room on the first floor of the Trippenhuis Building, Amsterdam, October 21-23, 2015.
“The ACTION project: dissemination”, session IV of the 5th International Conference on Neonatal and pediatric laboratory medicine, within the 11th International Workshop on Neonatology – From the womb to the adult, Cagliari, Italy, October 26-31, 2015. It had more than 1000 participants and was accompanied by many national and international institutions and scientific societies (Italian Ministry of Health and UNICEF, just to name a few).
Early career scientists workshop on Aggression – Nijmegen, The Netherlands, October 31, November 1, 2016. The event was organized following joint efforts from 4 European projects: ACTION, Aggressotype, Matrics, and FemNAT-CD, and was attended by PhD students and postdoc researchers from all four projects.
“Session: childhood aggression” and “Session: childhood aggression II” at the 47th Annual Meeting of the Behavior Genetics Association, Oslo, Norway, June 28 – July 1, 2017. The ACTION joined forces with EAGLE (EArly Genetics and Lifecourse Epidemiology), organized a symposium made up by two sessions to present the project results
Vu University Amsterdam, October 30-31, 2017. This meeting was associated to a Workshop on Aggression studies dedicated to junior researchers which has been held on October 30th. The event started with talks on aggression-related comorbidity and social topics and the afternoon session was devoted to Biomarkers and (epi)genetics topics.
“Session: Family and Molecular Genetic Studies Provide Insights into ADHD, its Comorbidities (Aggression And Auto-Immune Disorders) and Symptomatology” at The 48th Annual Meeting of the Behavior Genetics Association, Boston, USA, June, 2018
Page with videos and photo galleries: http://www.action-euproject.eu/content/videos-and-interviews-action-final-dissemination-meeting . Photo galleries and Albums are available on Facebook.
Objective 7.3: Inform stakeholders and general public about the project PresentACTION
Some actions performed within previous task (7.2) have been conducted in order to satisfy also the objectives of this task (7.3): e.g. Newsletter exploitation, organization of Workshops and training events. Other actions are also described in the next task (7.4)
Interactive tools. The main outcome of task 7.3 regards the creation of interactive tools in order to show to stakeholders and general public, the comorbidities of child aggression with other childhood psychopathologies. These interactive tools are based on the several results of the ACTION consortium that were published or in press. These tools are also based on several studies and backgrounds of the consortium's partners. These tool have been promoted through publications, abstracts, posters, lectures, newsletters. (see previous technical reports for details).
A first interactive tool, Comorbidity of child aggression tool (based on M. Bartels, et al., “Childhood aggression and the co-occurrence of behavioural and emotional problems: results across ages 3–16 years from multiple raters in six cohorts in the EU-ACTION project”, European Child {&} Adolescent Psychiatry, 2018) shows comorbidities of child aggression, displaying results from questionnaires obtained filled by multiple raters, in six cohorts in the EU-ACTION project”. It can be browsed by questionnaire (ATAC - CBCL - DCB - SDQ - MSNI) or by psychopathology and it is available at this url:
http://www.action-euproject.eu/ComorbidityChildAggression
A second interactive tool, “Teacher Ratings Child Aggression tool”, (based on A. Whipp et al., 2019, “Teacher-rated aggression and co-occurring behaviors among schoolchildren”, paper under preparation) has the overall objective of characterize the levels and associations of aggressive behavior and co-occurring behaviors in the school setting, by using teacher ratings of children at ages 7–14 from population based cohorts of children from Finland, the Netherlands, and the UK.
It is made up by two parts and each of the two parts, similarly to the previous tool, can be browsed by questionnaire (ATAC - CBCL - DCB - SDQ - MSNI) or by psychopathology:
http://www.action-euproject.eu/TeacherRatingsChildAggression
http://www.action-euproject.eu/TeacherRatingsChildAggressionCorrelation
Media coverage targeting the general public. General public is another target of the objective 7.3. During these years the general public has been targeted through the official website, the social media channels (Youtube in particular). The public has been targeted also through articles in online magazines quoting the project and announcing events related to ACTION project. E.g.:
25/10/2018 - Il mondo della Neonatologia a Cagliari, Andrea Congia, UnicaRadio.it
25/10/2018 - Il mondo della Neonatologia a Cagliari, Copenaghenhouse.it
25/10/2018 - A Cagliari workshop internazionale sulla neonatologia, Copenaghenhouse.it
08/05/2017 - Aggression aetiology unveiled, new treatments on the way - Cordis website (available in 6 languages, ACTION is quoted at the end of the article)
27/10/2015 - Press release and and Interview to the Italian tv Videolina
20/10/2015 - 28-minutes audio interview with Prof. Robert Plomin for BBC Radio 4's program - The life Scientific, BBC Radio4
2015 – 40-minutes audio interview with Prof. Robert Plomin - The Guardian newspaper
13/05/2014 - Start of FP7 project ACTION: a study into the complex gene-environment interplay of children with externalizing behavior problems, Emgo.nl
2014 - ACTION - Aggression in children, Healthcompetence.com
2014 - ACTION - Aggression in children, scholarshipdb.net
Objective 7.4: Inform scientists about the project CommunicACTION
Some actions performed within previous tasks (7.2 7.3) have been conducted in order to satisfy also the objectives of this task (7.4): e.g. Newsletter exploitation, organization of Workshops and training events, Release of Interactive tools
The consortium published more than 90 scientific articles (Journals and conference proceedings), Open access has been ensured) focused or related to the ACTION project.
The complete list is available in the ACTION website at the url: http://www.action-euproject.eu/biblio
The consortium participated to a series of symposia at conferences, workshops ,summer schools and scientific meetings. These activities are a key aspect of dissemination and exploitation efforts undertaken within ACTION Work Package 7. The major objectives of these activities are the following:
To ensure a clear understanding of the project and the overall objective, effective knowledge transfer, an integrated approach;
To communicate and share the project’s results and achievements with other researchers;
To receive feedback by professional communities;
To increase the partner network of stakeholders;
To ensure information of and influence on media users and to inform politicians, policy makers, the general public and target groups about the benefits that the ACTION Project can provide;
To utilize the feedback to improve the quality of project results;
To collect new ideas and identify opportunities for further collaborations and funding either as parallel or follow up project activities.
The following list reports about lectures and presentations given by the consortium staff within events managed by other organizations, while the events (workshops and conferences, sessions within workshops) organized or co-organized by the ACTION consortium are listed in the Objective 7.2 of this document and the related presentations (since each event comprises several talks by the ACTION staff members) are better detailed in the Deliverable 7.2.
Objective 7.5: Creation of an updated list of representatives of relevant stakeholders and en-user groups in an interactive stakeholders platform / T7.5 RealizACTION
The consortium has progressively created a liaison database of researchers and stakeholders, with the purposes described in the DoW and in the previous objectives. The consortium organized and participated to a series of symposia at conferences, summer schools and scientific meetings. A selection of contacts comprises :
- Liaison and synergies with other EU funded consortia
Aggressotype consortium
Matrics consortium
FemNAT-CD consortium
CAPICE consortium
- Organizations and institutions
Netherlands Institute for the Study of Crime and Law Enforcement (NSCR)
Union of European Neonatal & Perinatal Societies (UENPS)
European Academy of Paediatric Societies (EAPS)
European Association Perinatal Medicine (EAPM)
Union of the Middle-Eastern Countries and Mediterranean Pediatric Societies (UMEMPS)
Italian Society of Pediatrics
Italian Society of Neonatology
Hellenic Society of Neonatology
SI-DOHAD, Italian Society Developmental Origins of Health and Disease
International Federation Clinical Chemistry – IFCC
Unicef
National Institutes of Health (NIH), U.S. Department of Health & Human Services
Behavior Genetics Association (BGA)
International Society Twin Studies (ISTS)
European Child and Adolescent Psychiatry (ECAP)
Consortium on Individual Development (CID)
Exploitable results: The European ACTION program will achieve its desired impact by ensuring that the project obtains the required visibility through targeted communication measures, and by thoroughly planning dissemination and exploitation of results. Communication efforts will target scientific and other stakeholders, patient groups, counselors, teachers and social workers, and public health service stakeholders. Core activities to maximize the long-term impact of ACTION will ensure that:
The public website www.action-euproject.eu will be kept online for at least 1 year after the project end. This deadline will be almost surely extended to year 2022.
Scientific outcomes will be widely disseminated;
Project results published in peer-reviewed journals also after the end of the project;
Open access to scientific publications will be provided as early as possible and will also be accomplished through the PubMed Central reference number (PMCID) which is a unique number assigned to a work that is posted to PubMed Central (PMC), a free digital archive of biomedical and life sciences journal literature at the U.S. National Institutes of Health;
Project visibility will be created among relevant target groups.
Target audience: In this Section we focus on the possible targets, in terms of public and private institutions, researchers and general public, which may take advantage from news and research products of the ACTION project, after the project end. Several targets include: Scientific community, researchers , higher education, patient organizations, sychiatrists, pediatricians, health care workers, parents, teachers, policy makers
Target’s potential needs related to the achieved results of the project: The ACTION program will advance globally excellent research, which holds the promise of theoretical, methodological, and empirical innovation. It also seeks to promote research based knowledge and insight which has societal, practical and policy relevance. The novel contribution of ACTION will come from the emphasis on: the interactions and interdependencies between different levels of influence on childhood aggression. Rather than treating them as independent, ACTION adopts an interdisciplinary approach to understanding the relationships between different levels of risk and protective factors, focuses on different population groups, and considers the intersections between them to inform various stakeholders and policy development.
Potential Impact:
Potential impact and the main dissemination activities and the exploitation of results.
Some very relevant conclusions can be made about the collection of biomaterials in a clinical setting and its implications for future research.
Firstly, the willingness of parents and children to collect biomaterials did receive more resistance than expected. To study this resistance an inventory among patients refusing participation was conducted. The results of this inventory could function as a guide to inform participants on topics that they deem sensitive as well as maximizing the participation rates in this type of research. Furthermore, during data collection the severity of psychiatric problems of the clinical population increased considerably, which also resulted in lower participation rates.
Although the realization of biological samples was more difficult than expected, there were several major advantages compared to studies focusing on clear-cut specific diagnoses as well as population-based samples. The standardized collection of biomaterials in a broad defined psychiatric sample are guaranteed to have significant problems according to multiple sources (parents school, referrer, and clinicians at the psychiatric facility), enabling generalization to children that actually experience significant impairment from their problem behaviors. In contrast, research in population registers is most of the time reliant on a distribution of problem behaviors, not actual impairment. Additionally, recruitment of children that are referred to psychiatric institutions are screened extensively, therefore very detailed information is available. Importantly, this information is most of the time far more extensive than information which is collected in conventional research (which mostly focuses on parent-reported questionnaires).
Furthermore, the infrastructure for data collection which has been set up thanks to ACTION is still intact, with more participants being recruited to this day. Additionally, the ACTION consortium also realized the contacts with experts on biomarkers and (epi)genetics, which means that future research on biological aspects in a psychiatric setting can be realized more easily than before.
The results with respect to guidelines and clinical practices on severe behavioral problems (SBPs) in children as perceived by research and practice experts, help closing the gap between evidence-based practice and practice-based evidence.
With regards to guidelines, both academics and clinicians share a range of concerns and critical needs around their applicability. Both indicate that familiarity with and implementation of guidelines are modest. Questions around how guidelines are defined and disseminated in practice are also highlighted. Both groups of participants indicate that guidelines should be better adjusted to clinical practices, by expanding on the pool of evidence-based methods included, offering more specific recommendations based on a range of risk factors (e.g. comorbidity, severity, family dynamics, wider systems), taking a multi-factorial approach to formulation and treatment, and increasing focus on systemic working. Merging experts’ and clinicians’ perspectives on guidelines has also indicated insightful gaps in implementation. To this end, our findings inform that more efforts should be directed at persuading commissioners to prioritize SBPs, investing in additional staff training in recommended evidence-based methods, securing national government/ health authorities support, formatting guidelines in a concise and user-friendly format, translating recommendations across countries, and annexing dissemination guidelines. An action plan on key critical needs offers a framework of needs and foreseen challenges in improving guidelines fitness-for-purpose that could be used by guideline development groups when revising current or developing new guidelines. The project also instills a momentum for further development of national or European frameworks for managing SBPs, promoting learning between countries and researchers and clinicians. With regard to treatment practices, our findings provide an initial summary of available prevention and intervention programs across Europe, emphasizing existing gaps in service provision which could further be addressed and explored. The findings also highlight potential barriers to accessing treatment. Finally, by gathering an overview of common and different opinions amongst clinicians concerning specific diagnostic and treatment methods or elements used, the study offers first-hand perspectives on clinical judgment around what intervention is being used for each family and each child and what is perceived as helpful for SBPs. This overview can further provide practice-based evidence to inform guidelines and policy development. From a research perspective, these projects emphasize the value of guidelines in enhancing a shared framework for conceptualizing SBPs to reduce research-practice gaps. The broad overview provided may also inform future more systematic evaluations of highlighted issues. Finally, it promotes sharing of information and expertise between countries.
We found a relation between the age at diagnosis of problem behaviors related to aggression and distal and established that a later diagnosis predicts a higher probability of having a criminal record after age 15 and a lower income as a young adult. Age at time of diagnosis did not affect grade scores. The effect of age at time of diagnosis on suicide attempts was not significant after correcting for multiple testing. Importantly, we showed that the predictive effects of the timing of when the diagnosis occurred on later income and criminal records are independent of effects of background variables that were included in the analyses (i.e. mother’s education and criminal records, neighborhood deprivation, family income, ADHD and ODD diagnoses). The effects were very similar for boys and girls. Our results point in the direction that an early diagnosis could provide the basis for intervention and treatment settings, which may influence the developmental trajectory into later criminality. While very few larger-scale studies have investigated the specific effects of timing of interventions for any neurodevelopmental or disruptive disorder, early intervention has been shown to be effective, especially in children with high initial risk. Our results also support that early intervention has a positive effect on long-term outcomes related to criminality. The effects are in line with the understanding that neural plasticity in the developing brain makes it particularly susceptible to a broad range of interventions. However, given the fact that our data did not contain information on the start of any treatment, some caution is warranted. For instance, it could also be possible that a later diagnosis is merely indicative of persistence, which in turn might be related to poorer outcomes.
Our the results indicate that it is important to monitor aggressive boys and girls not only for future externalizing problems, but also for future internalizing problems. Prevention of specific outcomes, however, may require different interventions. Regarding medication, in young men who received second generation antipsychotics, rates of violent reoffending were lower during periods when individuals received this compared with periods when they did not. Similarly; among released prisoners in Sweden, rates of violent reoffending were lower during periods when individuals were dispensed antipsychotics, psychostimulants, and drugs for addictive disorders, compared with periods in which they were not dispensed these medications. This is valuable information, although further research is needed to understand the causal nature of this association. Finally, SSRI use is associated with elevated rates of violent crime, which remain throughout treatment and persist after discontinuation. Despite the rarity of the outcome, our findings suggest the need for clinical awareness of the risk for violent crime while taking SSRIs, and for provision of information to present and future users of this medication.
The genetic architecture, including the longitudinal genetic architecture, of aggressive behavior in childhood is comparable between countries even with the use of different instruments. An important message for parents, teachers and counselors is that aggressive behavior is heritable and the genes are a main source for differences between children in aggressive behavior. Given the stability of genetic effects it is possible to use phenotypic data collected by the different partners across a large age range. Psychiatric, social, and parenting interventions for Antisocial personality disorder (ASPD) prevention should focus on children and adolescents with high aggression levels, with an aim to gather information from multiple informants.
Aggression co-occurs with almost all other behavioral and social problems. More specifically, aggression co-occurs with oppositional and ADHD-related problems, and at later ages with rule-breaking. In addition to the high correlations of aggression with externalizing problems, we also observed substantial associations with anxiety–depression and other internalizing symptoms. Both for externalizing and internalizing problems, the patterns of co-occurrence were largely gender and rater independent, and were similar even when aggression and the other psychopathologies were assessed by different assessment instruments. Also, there were large similarities in co-occurrence patterns across countries in the Northern part of Europe. The finding that aggression co-occurs with nearly all other behavioral, emotional and social problems during childhood should put aggression in the center of attention, both for research and clinical perspectives.
When causes of differences in aggression during childhood are better understood, this information aids in the development of prevention and intervention strategies. Across definitions, conclusions as to whether different AGG instruments measure the same construct differ. Item content suggests limited overlap, whereas genetic analyses suggest shared etiology among instruments. Whether researchers regard agreement between instruments as satisfactorily depends on their application. It is highly recommended to consider multiple metrics of similarity to decide whether differ measures assess the same. By leveraging a genetically informative design and several commonly used instruments, we attempted to provide a holistic perspective on nuances involved in measurement of AGG in childhood.
The largest genetic association study to date of childhood AGG identified no genome-wide significant associations at SNP level, but identified two genes that are significantly associated with AGG in a gene-based analysis: PCDH7 (protocadherin 7) on chromosome 4p15.1 and ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase 3) located on 1p34.1. There was significant heritability based on single nucleotide polymorphisms. Genetic correlations were high and positive for a series of disorders and behaviors including ADHD, Bipolar disorder, Loneliness, Major depression, Risk tolerance, and Smoking. Genetic correlations were high and negative for educational attainment and age of first child. Stratified analyses revealed significant differential genetic correlations between AGG and other behavioral, emotional and social problems, and life outcomes; highlighting the importance of combining information from multiple sources to get the most complete overview of the genetic etiology of childhood AGG.
Within Action, we have generated a large amount of biomarker and ‘omics data to study childhood aggression. We developed and piloted extensive protocols for sample collection, i.e. urine and buccal-cell collection in children, which were made available for the scientific community (http://www.action-euproject.eu/content/data-protocols) for future projects.
In ACTION we report the first findings of the genetic association studies, the epigenetic projects and the metabolomics studies. All genome-wide results and all products from the metabolomics projects are made available to the scientific community, to not only aid in the understanding of childhood aggression, but also to study the links with adult outcomes through techniques that do not require the raw data, but can take summary statistics to estimate genetic correlations, construct polygenetic scores for risk prediction, based on aggression and attention problems discovery studies and on the extensive database for SNP-metabolomics associations, because we have also undertaken a large-scale investigation of the genetic architecture of blood metabolites. As part of this undertaking we created a database of all published (2008-2018) metabolite-SNP associations, which includes characterization and classification of all metabolites. This database has been made publicly available (http://bbmri.researchlumc.nl/atlas/#data). Overall, ACTION provides tools for future scientific research, ranging from potential amino acid intervention studies (protocol), to polygenetic prediction or Mendelian Randomization studies (SNP-metabolite associations) and multi-omics investigations (ACTION ‘omics data).
For Action, a steroid hormone platform to be used for urine samples, was developed which offers great opportunities to the scientific community to explore the potential relationship of hormones and (subtypes of) aggression and other behaviors and behavioral problems in children.
The extensive findings in Action reflect collaborative efforts to integrate genetic, genomic and multilevel environmental methods to better characterise aggression in childhood. Details of the advanced statistical modelling and study results and implications have been communicated internationally in order to widen the impact and public understanding of the genetic and environmental antecedents to aggression. Through high impact publications, media coverage and invited lectures and conferences, we have collectively raised the research profile of childhood aggression and demonstrated the utility of large genetically sensitive collaborative efforts in studying its etiology. Crucially, our multi-method approach to investigate GxE and rGE using twin, polygenic score, epigenetic and atheoretical machine learning has produced exciting preliminary results which are currently being replicated in the ACTION cohorts and will promote future dissemination activities even after the end of the ACTION grant period.
In collaborative projects, it is not uncommon that research groups use different instruments to assess a trait. We examined convergence between different instruments to assess aggressive behaviors, focusing on item content, clinical concordance, correlation, and genetic overlap. The findings revealed that agreement between different measures of childhood aggressive behaviors depended on the metric of agreement under consideration (i.e. item content, clinical concordance, correlation, underlying genetics). Overlap between the item content of the aggressive behavior measures was absent to moderate. Concordance on who received a score above the clinical cut-off was very weak to weak. Associations between the different measures of aggressive behavior based on continuous scores, while correcting for skewness, yielded higher agreement (i.e. moderate to strong) than clinical cut-off scores. Genetic correlations ranged from weak to very strong, which generally indicated high overlap in underlying genetics between the different measures of aggressive behavior. From this study, two patterns emerged that may impact future practice and research. First, higher convergence between instruments for correlation than for clinical concordance suggests that decision-making based on continuous scores instead of clinical cut-offs may improve reliability. Second, the generally high genetic overlap suggests that the same genetic factors influence the different measures of aggressive behavior, despite lower convergence in item content, clinical concordance, and correlation. Therefore, the findings indicate that different measures of aggressive behavior can readily be combined in future collaboration studies on the genetics of childhood aggressive behavior.
We also provided a harmonized aggression score that facilitates future collaborations to further investigate the etiology of childhood aggression. This work allowed for a more generalized statement about aggression in school-aged children from the Netherlands, Finland, Sweden, and the United Kingdom. The harmonized aggression score permitted analyses to be conducted in a large international sample in which all participants had a comparable score. The harmonized aggression score is available to ACTION partners for future joint analyses. Moreover, the reference panel created for this study may facilitate future linking and scaling of phenotype scores across and beyond the consortium. The integrated data analysis framework discussed in the study provides promise for future multi-cohort collaborations to investigate behavioral or psychological phenotypes. Often different cohorts assess traits using different items which introduces measurement heterogeneity. The method presented in this study should remove this specific type of heterogeneity and consequently benefit future joint analyses of multiple studies.
Both genetic and environmental factors are important contributors to individual differences in childhood aggression and a comprehensive understanding of the etiology of childhood aggression requires taking both genetic and environmental factors into account, as well as examining the possible interplay between them. A socio-ecological framework was considered that acknowledges the importance of biological and psychological systems at the individual level together with factors at distal levels including parents and the family, the community, and finally the geopolitical level.
At the distal level, there is a negative association between national-level policies for child and adolescent mental health (CAMH) and adolescent aggressive behavior (N = 172,829, from 30 European countries). This indicated lower aggressive behavior in countries with more CAMH policies. National-level policies in this study consisted of measures of availability of epidemiological data, the number of CAMH facilities, investment in family benefits, and investment in education. We obtained information on these policies from renowned statistical institutes (e.g. Eurostat, 2016; Institute for Health Metrics and Evaluation, 2018; OECD Social Policy Division, 2016). This study was the first to examine the association between CAMH policies and adolescent aggressive behavior, and revealed the potential of CAMH policies to function as large-scale prevention to ameliorate adolescent aggressive behavior. Furthermore, this study provided directions for future research to better understand the mechanisms underlying this association.
On a more proximal level, we aimed to predict childhood aggression with a wide range of family- and individual-level predictor variables. The outcome measure was physical overt aggression, assessed around age 9 and harmonized across multiple European cohorts. Application of rigorous methods that permitted simultaneous analyses of all included predictor variables yielded a multi-group model that accounted for interactions with sex and cohort and confirmed linear main effects of variables measuring non-physical aggression, ADHD and conduct disorder related behaviors, maternal smoking during pregnancy, parenting, and proportion of life events. The results provided a set of easily observable behaviors that might help to identify children more likely to become aggressive. Simultaneous modelling of variables yielded insight in their importance when also taking other variables into account. An additional benefit from the rigorous methods was that they allowed for testing different types of predictor effects (i.e. linear, non-linear, interaction) without having to pre-specify them. These methods may provide tools for future (collaborative) research projects that aim to predict a trait using a wide range of variables .
The contribution of genetic and environmental factors to individual differences for a trait may vary as a result of different backgrounds (i.e. gene-environment interaction), such as levels of socioeconomic status (SES). We examined the moderating effect of SES on the genetic architecture of childhood aggression in large samples of 7-year-old twins SES moderated the contribution of genetic and environmental factors. The contribution of genetic factors was similar across SES strata, the contribution of shared environmental factors was lower, and the contribution of nonshared environmental factors was higher for children from a high SES background compared to children from a low or medium SES background. This pattern was similar for children from the Netherlands and the United Kingdom.
This finding indicates that children from low or medium SES backgrounds would benefit more from treatment to improve shared environmental factors. Examples of such shared environmental factors are healthy family functioning, less parental stress, housing quality, or positive school attachment. To some extent, policies for child and adolescent mental health such as investment in family benefits and investment in education attempt to already do so through alleviation of factors associated with low SES. The results suggest that such early prevention measures on the shared environment indeed may benefit children at risk for heightened levels of aggression.
The literature synthesis of 72 systematic reviews and meta-analyses on treatment effectiveness for childhood aggression revealed that effect sizes were mostly absent or small for universal and selective prevention, and mostly small to medium for indicated prevention and intervention. For moderators including child age, child gender, implementation to individuals or groups, person implementing the treatment, different treatment programs, and session related factors or treatment intensity, these effect sizes were invariant in a majority of studies. Moderator effects were mixed for socioeconomic status, type of treatment, informant, and research quality. The only two significant moderators comprised of pre-treatment levels of aggression and parental involvement. Treatment effectiveness was higher for children with higher levels of aggression before treatment and when parents were involved in the treatment.
A first pattern that emerged from the literature synthesis was that children with sub-clinical levels of aggression appeared to benefit as much from treatment as children with clinical levels of childhood aggression. Secondly, results revealed that based on existing research it is not yet possible to distinguish subgroups of children that would benefit more from treatment for aggression than others. The positive moderating effect of parental involvement on treatment effectiveness for childhood aggression suggests that an opportunity for future research may be to focus more on parental influences as possible moderators of treatment effectiveness. These findings provide additional support that it would be beneficial to screen for parental psychiatric disorders for the treatment of childhood behavior problems .
The wealth of available data and expertise within ACTION permitted examination of new research questions. The studies provided tools for the assessment of childhood aggression (and other traits) in future collaborative research projects. Assessing whether children score above or below a clinical cut-off for inclusion for treatment may lead to children to be excluded from treatment from which they would benefit as much as children who score above a clinical cut-off. Moreover, measures of aggressive behavior agree only to a small extent on which children display clinical levels of aggression, which may cause children to miss out on treatment not because of their level of aggression, but because of the measure selected to assess their aggression. A harmonized factor score of aggression may facilitate future collaborative research projects. In addition, the research in WP6 highlighted the complexities in the etiology of childhood aggression. Childhood aggression is found to be associated with a broad range of factors, from country-level policies to more proximal factors as the family environment, and individual level factors such as behavior and genetics, which may also interact.
Although the influence of the broad range of factors we considered on childhood aggression adds complexity to the etiology of childhood aggression, it also provides opportunities to improve prevention and intervention strategies for childhood aggression. For example, inclusion of parental characteristics (i.e. a family based approach) in diagnosis and treatment might improve treatment effectiveness for childhood aggression. Additionally, more policies for child and adolescent mental health were associated with lower levels of aggressive behaviors, which suggests merit in employment of policies as early prevention efforts. Moreover, differences in etiology of aggression as a result of socioeconomic background highlight that it is promising to distinguish subgroups of children more likely to develop childhood aggression and children more likely to benefit from treatment. The research contributes to previous work to advance our understanding of assessment, etiology, and treatment of childhood aggression and provided directions for future research working towards a more personalized approach to childhood aggression.
Action has been very involved in all dissemination activities with a very accessible website, movies newsletters and interviews on Youtube channels, and a great set of interactive tools.
Digital presence:
Website: www.action-euproject.eu
Youtube channel: https://www.youtube.com/watch?v=j0Ev8h1ARkE
Hahshtag on Facebook and Twitter: #actionproject
Comorbidity of child aggression tools:
http://www.action-euproject.eu/ComorbidityChildAggression
Total number of interactive charts: 25
- Teacher Ratings Child Aggression tool
http://www.action-euproject.eu/TeacherRatingsChildAggression
http://www.action-euproject.eu/TeacherRatingsChildAggressionCorrelation
Total number of interactive charts: 35
Videos, interviews :
o ACTION official video: https://www.youtube.com/watch?v=j0Ev8h1ARkE
Multiple Videos from the ACTION final meeting on aggression studies, May 6-8, 2019 Pula, Cagliari, Sardinia, Italy
- Childhood aggression and its comorbidities: overview of the dissemination meeting: https://youtu.be/fFPXU9Hmrpw
- Prof. Meike Bartels: main results of the ACTION project - Childhood aggression and its comorbidities: https://youtu.be/PJBocGFurBo
- Prof. Robert Vermeiren: clinical implications of the ACTION project - Childhood aggression: https://youtu.be/_jdtr5I-mio
- Prof. Gitta Lubke: Predicting childhood aggression; mining large datasets followed by confirmatory modeling: https://youtu.be/qKcyqWIh-7Q
- Dr. Jenny van Dongen: Epigenome-wide association study meta-analysis of aggressive behavior: https://youtu.be/y7HUrl9oZow
- Hill Ip, PhD candidate: Genome-Wide Meta-Analysis of Aggressive Behavior across Rater, Instrument and Age: https://youtu.be/wpDax_Hiq9I
- Dr. Margherita Malanchini: aggressive behavior in childhood and the role of smoking during pregnancy: https://youtu.be/w2oQ6E0uncc
- Yayouk Willems, PhD Candidate: Environmental and Genetic contributions to low self-control: https://youtu.be/lg21SEugGeM
- Peter Roetman, PhD Candidate: aggression in clinical practice: https://youtu.be/Xz8Qs9ZYxzI
- Andrea Allegrini, PhD Candidate: Multi-trait polygenic prediction of broad aggression in adolescence: https://youtu.be/DUApzodNyEg
- Dr. Lucia Colodro Conde: a twin study of oppositional defiant disorder behaviors using data from Australia and The Netherlands: https://youtu.be/CMdw6uxfBRs
- Fiona Hagenbeek, PhD Candidate: Urinary biomarkers and metabolites of childhood aggression: https://youtu.be/74zK-zD9-6I
- Anne Hendriks, PhD Candidate: What are the implications of five years of research on childhood aggression? https://youtu.be/22ZSXXmTDNs
- Alyce Whipp, PhD Candidate: Teacher-rated aggression and co-occurring behaviors: https://youtu.be/bf59AxosxII
- Roberta Pintus: Lipids and aggression: what is going on?: https://youtu.be/3Uvaa2YQNwQ
- Childhood aggression and its comorbidities: concluding remarks and award ceremony (best talk given by PhD candidates: Fiona Hagenbeek - Urinary biomarkers and metabolites of childhood aggression): https://youtu.be/PekNHUZLDKU
o Videos from the "14th International Workshop on Neonatology", Cagliari, Italy, October 24-27, 2018.
- Interview with Prof. Vassilios Fanos - session for early researchers within the 14th Workshop on Neonatology:https://youtu.be/jHD1rqnryaA
- Interview with Dr. Mirko Manchia - Clinical implications of the ACTION project
https://youtu.be/RJCOCYLH_8Q
- Interview with Dr. Matteo Mauri - Role of the University of Cagliari in ACTION, dissemination and main results of the ACTION project: https://youtu.be/j1vrt2eW6QM
o Videos from the "13th International Workshop on Neonatology", Cagliari, Italy, October 25-28, 2017.
- Interview with Prof. Dorret Boomsma - Main results of the ACTION project: https://youtu.be/6jmwszJXMLM
- Interview with Prof. Fanos - The ACTION project's goals: https://youtu.be/PL5NiepwWto
- Interview with Dr. Mirko Manchia - Risk factors for aggression: https://youtu.be/ewbbh-UoHU4
- Interview with Prof. Luigi Atzori - Metabolomic approach and aggression, in Italian language: https://youtu.be/-9SOaTicUVM
- Other interviews and videos (11 in total) sponsored by ACTION: official Youtube page of the event
o Other Videos/Interviews (last access 2016)
- Project Kick-off meeting: https://youtu.be/c5b_4dtg8nA
- Interview with Robert Plomin in Times Education Supplement - Genetics and Education: https://youtu.be/-4VCISnwE-k
- "A new Scientific Humanism: Seven Chapters on Narrative Medicine and Precision Medicine", Florence 18-19 February 2016: http://www.action-euproject.eu/content/new-scientific-humanism-seven-chapters-narrative-medicine-and-precision-medicine-florence-18
- Webinar at NIH - National Institutes of Health, U.S. Department of Health & Human Services: Application Metabolomics to Provide Pediatric Biomarkers: https://youtu.be/URVBX57qUdI
- Interview with prof. Vassilios Fanos (Italian TV Videolina): http://www.action-euproject.eu/content/interview-prof-vassilios-fanos-italian-tv
Newsletters (all available on website)
ACTION Newsletter N°1, December 2015;
ACTION Newsletter N°2, August 2016;
ACTION Newsletter N°3, July 2017 ;
ACTION Newsletter NTR SPECIAL ISSUE, September 2017;
ACTION Newsletter N°4, March 2018;
ACTION Newsletter N°5, ACTION Final meeting invitation, February 2019 (not available for download)
List of Websites:
public website: http://www.action-euproject.eu/
contact details: n.stroo@vu.nl
Childhood aggression can be described in many ways, by multiple people who are familiar with a child and by many instruments and methods. The Action consortium investigated causes and consequences of aggressive behavior in childhood, as assessed from multiple perspectives, in order to improve the understanding of the genetic and non-genetic etiology of aggression in children and to inform on the development of prevention and treatment strategies. Action brought together data from twin cohorts and genotyped cohorts to estimate and identify the causes of variation in aggression, to disentangle (epi)genetic and environmental effects, to inform on comorbid disorders, to investigate biomarkers and to gain insight in the predictive power of pediatric aggression for later outcomes.
We showed that aggressive behavior in childhood is associated with negative behavioral outcomes and that that it predicts emotional and cognitive problems. We established that early detection of aggressive problems and access to treatment results in significantly lower crime records and higher average income. We also showed that across development and environmental context, aggressive problems do not occur in isolation: the majority of children who are aggressive also suffer from attention problems, anxiety and depression. There is a robust negative association between aggression and academic performance in compulsory education. The large comorbidity of aggression with other childhood psychopathology was also indicated by clinicians, social workers, parents and teachers of children as one of the major difficulties in treating this group.
To obtain an understanding of why children differ from each in aggression, ACTION established an international consortium among partners with access to large childhood cohorts from Finland, Sweden, UK, the Netherlands, Italy and Australia. The longitudinal data established the substantial heritability of aggression across age and gender. The stability of aggression across age is large and due to the constant genetic influences on aggression. Next, Action extended the consortium to include many research groups from around the globe, in meta-analysis projects that focus on identification of genetic markers and differentially methylated regions of aggression and one of its major comorbid conditions, i.e. attention problems.
The genome-wide epigenetic results based on a meta-analysis of 14,434 peripheral blood samples across all ages and 2425 cord blood samples indicated 48 top CpGs for which we performed enrichment analyses. The top ten most strongly enriched traits indicate that CpGs associated with aggressive behavior overlap with CpGs previously associated with smoking, maternal smoking, and other chemical exposures (e.g. perinatal exposure to perinatal polychlorinated biphenyls (PCBs) and polychlorinated dibenzofurans (PCDFs)). Further overlap includes CpGs associated with alcohol use, cognitive function, educational attainment, ageing, and metabolic traits.
The first large genetic association meta-analysis (GWAMA) of childhood aggression identified two genes that are significantly associated with AGG in a gene-based analysis: PCDH7 (protocadherin 7) on chromosome 4p15.1 which has the highest expression in the brain and spleen and ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase 3) located on 1p34.1 which is most highly expressed in testis and fat tissue. There was significant heritability based on single nucleotide polymorphisms. Genetic correlations were high and positive for a series of disorders and behaviors including ADHD, Bipolar disorder, Loneliness, Major depression , Risk tolerance, and Smoking. Genetic correlations were high and negative for educational attainment and age of first child.
Project Context and Objectives:
Within Action, we worked on a set of interrelated objectives: Within Clinical Epidemiology we aimed to 1) gain insight in the treatment and diagnostic practices in Europe and identify the critical needs for stakeholders; 2) quantify and identify the long-term societal outcomes of childhood aggressive behavior; and 3) develop long-term positive and negative trajectories from early childhood to adulthood for subtypes of aggression. In Genetic Epidemiology, we aimed to 1) unravel the genetic longitudinal architecture of aggression and estimate heritability; 2) resolve patterns of comorbidity of aggression and emotional and behavioral problems; 3) detect genomic regions for aggression in children in genome-wide association (GWA) studies; and 4) identify differentially methylated regions related to aggression, also in a genome-wide approach. In Gene-Environment Interplay we focused on environment in order to: 1) identify environmental risks that have effects on the development of aggression independent of genetics; 2) identify environmental effects on the development of aggression that depend on genetic propensities (GxE interaction); 3) identify genetically driven experiences associated with aggression in which children select, modify or create environments correlated with their genetic propensities (GE correlation); 4) apply methods for SNP heritability and polygenetic predictors; and 5) extend analyses of epigenetics to investigate GE interplay. For the work on Metabolomics & Biomarkers we collected buccal swabs and urine samples in children from clinical and population settings. Action aimed to 1) identify biomarkers to diagnose aggression; 2) identify biomarkers to differentiate between sub-groups of aggression; and to 3) identify treatment options.
The work realized in these four large epidemiologically oriented approaches informed on treatment and prevention strategies and lead to 1) development of an evidence-based model on effectiveness of existing preventative and treatment strategies; 2) integration of insights in a comprehensive theoretical framework; 3) development of risk assessment charts and guidelines to improve clinical decision making; and the assessment of this framework and risk assessment charts with respect to ethical, legal and social implications (ELSI). Within our work package Dissemination and stakeholder interaction we collaborated to 1) provide a recognizable and global project identity; 2) update participants and participating agencies; 3) inform stakeholders and general public; 4) inform the scientific community about the project; and 5) create a list of stakeholders and end-user groups.
In our clinical epidemiology study of experts and clinicians from 24 countries of practices in Europe, experts supported developing European guidelines and emphasized a need to develop Evidence Based Methods. Clinicians emphasized limited usefulness of guidelines for clinical practice. Therapeutic alliance, focus on positives, improving parent-child interactions, and collaboration with schools and teachers were stressed to improve prevention and treatment of severe behavioral problems. Comorbidity was identified as a major theme. Comorbidity was the focus of one of two of our large empirical study across the childhood cohorts in ACTION. We showed that co-occurrence of aggression and other behavioral and emotional problems is substantial, and that this pattern strongly emerges independent the person who rates the child’s behavior (parent or teacher). There is an especially high degree of co-occurrence of aggression with attention problems, rule breaking, oppositional behavior, and substantial associations are also seen with depression, anxiety, and autism. The ACTION website has a set of interactive tools www.action-euproject.eu/ComorbidityChildAggression and
http://www.action-euproject.eu/content/new-interactive-tool-released showing the comorbidities of aggression with a wide range of childhood problems, across ages, raters and instruments.
A record linkage study in Sweden of long-term societal outcomes of childhood aggression revealed how long-term positive and negative trajectories from early childhood to adulthood develop for aggression subtypes; how early detection of aggressive problems and access to treatment results in significantly lower crime records and higher average income. Suicide risk might also decrease.
A systematic literature review of effectiveness of prevention and intervention described effect sizes across treatments and moderator effects. For prevention and intervention, effects were small or medium. For moderators, effects were mixed. Moderators with a positive effect on treatment effectiveness for childhood aggression were pre-test levels of aggression, individual implementation, and parental involvement. Future research should distinguish between targets of treatment, focus on individual differences between children in treatment effectiveness and parental influences. A second systemic review summarized the current knowledge on the genetic basis of human aggression. As part of this project, we introduced an automated tool for the selection of literature based on supervised machine learning. This review summarizes all heritability and genetic association studies of aggression and concluded that power gains for GWASs on aggression need to be realized by increasing sample sizes, by combining data from different cohorts and by meta-analyzing genetically related outcomes.
Genetics plays an important role in explaining the individual differences in childhood aggression with heritability estimates of >50% and much smaller contributions of cultural inheritance and shared environment and genetic factors also are the main drivers of stability in aggressive behavior. A phenotype reference panel was realized, in which aggression was assessed on the same children by multiple instruments as used in the ACTION cohorts. We realized a survey study that included the CBCL (used in Holland and Sweden), the SDQ (used in UK), and the A-TAC (used in Sweden) for 2508 twins. Our results demonstrated heritability and high genetic overlap, suggesting that the underlying construct of childhood AGG is consistent across measures. This finding holds promise for gene finding studies.
To identify genetic variants, two large genetic association meta-analyses (GWAMA) for Aggression and ADHD were finalized, for which new meta-analytic approaches were developed. The GWAMAs include data from 30 cohorts, extending the Action consortium significantly to other childhood cohorts, with over 500.000 data points from multiple raters. Two genes are significantly associated with AGG in the gene-based analysis: PCDH7 (protocadherin 7) located on 4p15.1 which has the highest expression in the brain and spleen and ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase 3) located on 1p34.1. It is most highly expressed in testis and fat tissue. While we find no genome-wide significant SNPs in the overall GWAMA, we find significant SNP heritability and genetic correlations for a wide range of outcomes: childhood AGG is substantially correlated with ADHD and adult psychiatric disorders, including Bipolar Disorder, Schizophrenia , and Major depression. Substantial correlations were also seen for loneliness, risk tolerance, and smoking. Genetic correlations were high and negative for educational attainment and age of first child. The GWAMAs provide important information for the broader psychiatric genetics community. Action will make the meta-analysis results available, so that polygenetic scores (PGS) can be obtained for prediction in independent cohorts. Equally important are results for research on modification of genetic influences by environment, either through considering the totality of such influences by modeling of genetic relatedness matrices, or though considering the interaction with PGS and environment.
To identify differentially methylated regions we also were able to contact and include multiple international cohorts. Two EWAS projects that carried out meta-analysis of internationally available data were completed: one included 3 cohorts with genome-wide epigenetic data and ADHD. Findings point at new candidate loci involved in immune and neuronal function. The second EWAS includes all cohorts with aggression phenotypes and genome-wide epigenetic data, again extending the consortium far beyond its original members. This EWAMA includes 20 cohorts with data for 14,438 peripheral blood samples (mean age at blood sampling from 4 to 70 years) and 2195 cord blood samples. The mean age at phenotypic assessment is 6 to 72 years and aggressive or conduct problems were assessed by multiple instruments. In 48 top CpGs from the meta-analysis we performed enrichment analyses in the EWAS atlas against all reported associations with diseases and environmental exposures. The top ten most strongly enriched traits indicate that CpGs associated with aggressive behavior overlap with CpGs previously associated with smoking (37 CpGs; corresponding to 77% of aggression-associated CpGs and 0.3% of CpGs that were previously associated with smoking), with other smoking traits (e.g. maternal smoking), other chemical exposures (e.g. perinatal exposure to perinatal polychlorinated biphenyls (PCBs) and polychlorinated dibenzofurans (PCDFs)). Further overlap includes CpGs associated with alcohol use, cognitive function, educational attainment, ageing, and metabolic traits. These results confirm the important role of that exposure to nicotine that was also established in the analyses of the twin cohort data.
Our study of environment and interplay of genes and environment shows that perinatal factors explained up to 5% of the variance in adolescent aggression. Maternal smoking during pregnancy predicted aggression at all ages independent of other perinatal and demographic factors. A multidimensional risk score based on home and school risks predicted 10% of adolescent aggression. The multidimensional risk factors of aggression spanned multiple domains of adolescents’ lives. Integrating environmental and polygenic predictors and assessing their predictive accuracy indicate that genotype-environment interaction and correlations are of very small effect size.
The study of biomarkers in urine has been done in a technical (n=20), a biochemical pilot (n=218) and in the main biomarker study (n= 1142 selected (on high and low aggression) twins and 186 clinical cases. Assessments were done for Creatinine, Density, Neopterin, C-peptide, Substance P, Oxidized DNA and RNA and on metabolomics platforms for Amines and Organic Acids platforms. We began with a review of the existing literature and identified 4 classes of potential biomarkers for human aggression: 1) lipids, 2) inflammation markers, 3) neurotransmitters and 4) hormones. With the work performed in Action we find suggestive evidence for the role of inflammation markers, lipids and amino acids involved in neuro-transmitter pathways in aggression. In children we identified metabolites of the cysteine metabolism pathway, involved in oxidative stress, as among the most promising aggression biomarkers. In a project in adults we found cytokines to be marginally significantly related to aggression differences in discordant MZ pairs and LDL cholesterol levels to be lower for the high scoring MZ twins as compared to their co-twin scoring low on aggression. Finally, the top 25% most associated metabolites in ACTION’s Main Study included multiple (essential) amino acids which are involved in neurotransmitter pathways, or may act as neurotransmitters themselves, such as the metabolites included in the cysteine metabolism pathway.
Research on the biological and genetic components of aggression needs to be balanced by attention to the context and environment in which children develop. ACTION contributes to a better understanding of aggression by examining children’s aggressive behavior in the context in which it develops. A child who is aggressive in the school context requires different treatment than one who is aggressive in the context of chronic stress associated with parental psychopathology. The type of intervention that is adaptive for children exposed to long-lasting economic hardship and socio-economic disadvantage may differ from youth who live in more economically advantaged environments. Polygenic scores (PGS ) based on genetic variants from the discovery GWA results can be applied, together with the results from the methylation, biomarker, environmental context analyses to build models predicting which children have a high risk for persistent aggression. This model can be used in future for stratification of high and low risk groups to investigate whether and which more intense treatments improve outcomes for high risk children.
The ACTION website is continuously updated and now also includes the Standard Operating Procedures for collaborative projects (http://www.action-euproject.eu/content/data-protocols ), publications, videos, interviews with (young) researchers and newsletters.
Our Scientific Advisory Board concluded “ACTION has delivered a thorough foundation for ongoing and future basic studies into childhood aggression especially by producing extensive and complete overviews on clinical practices and on genetic influences on aggression. The consortium has invested in innovative methodology for literature mining and established a large collaboration of cohort and clinical studies into childhood aggression in Finland, Sweden, UK, The Netherlands, Italy and Australia. A diversity of highly relevant aspects of this topic was studied in these large datasets such as the correlated morbidities that coincide with aggression in children, parental factors, the stability of genetic influences over time (very stable in longitudinal studies), genetic overlap of aggression with other disorders (such as ADHD) and longstanding questions such as the role of maternal smoking during pregnancy. Scientifically the project is a great success with 90 papers published and many more expected to come in the next year on the basis of the novel data collected. Data in these cohorts have been harmonized and brought together for basic studies into genetics and early and later environmental factors pivotal to the development of childhood aggression. From a public health point of view, the EU was right in supporting this multidisciplinary project which is and will be producing robust evidence for the insights generated”.
Project Results:
Below, we report for our Action project [Aggression in Children: Unraveling gene-environment interplay to inform Treatment and InterventiON strategies] the main outcomes for all objectives, as initially stated in our application, to address questions regarding paediatric conduct disorders and aggressive behavior. Our Action project had as its broad aim to improve the understanding of the genetic and non-genetic etiology of aggression in children, with the intention to inform the development of novel prevention and treatment strategies. We set out to unravel in large twin and genotyped cohorts the causes of variation in aggression, disentangling (epi)genetic and environmental effects and their interplay with a focus on critical developmental periods, gender, and comorbid disorders; to examine metabolomics profiles of aggressive behavior and gain insight regarding the predictive power of pediatric aggression for adult outcome variables.
Below, the results for each of the objectives of the project are summarized. These objectives were part of a series of interrelated work packages, that had their focus on Clinical Epidemiology, Genetic Epidemiology, Gene-environment interplay, Biomarkers, Societal implications and Dissemination.
Results for Objectives from our work package Clinical Epidemiology.
Objective 2.1: Select for sample collection for epigenetic and biomarkers studies children (aged 6-12 year) from clinical setting and collect buccal swabs and urine
Sample collection was realized in 189 children. Based on our empirical experiences, we addressed the difficulty of including participants from clinical settings. In projects such as ours, researchers can be faced with decreases in the number of children admitted to clinics, because of reorganization of child and adolescent psychiatry policies, such as happened in the Netherlands. Related to this, the relative number of children with acute mental health problems increased and due to the reorganization of Dutch child and adolescent psychiatry and severe budget cuts, institutions had to decline participation. We have analyzed the whole process by approaching parents of children under these conditions. Based on a quantitative analysis of (non-)participants, the following changes in the recruitment procedure have been implemented: the content of telephonic interviews was adjusted to offer options that make participation easier, e.g. the possibility of a researcher making a house visit instead of the child having to visit the clinic or the possibility that a researcher instead of a parent collects the urine and buccal swaps of children who are spending the night at a clinic. A more personal approach was adopted to explain the study. A pilot was done in which parents who indicated interest in participating during the first contact, receive study details in a face-to face setting. We consider this better than mailing study folders and informed consent forms. Reasons given by parents largely relate to the stress experienced when entering a mental health facility. They feel too exhausted to invest in research participation, although they consider this of great importance.
Objective 2.2: Gain insight into the current treatment practices (guidelines, interventions and working components) for severe behavioral problems (SBPs) in children in Europe and identify the critical needs for stakeholders.
This objective was reached through conducting qualitative interviews in Europe to investigate the availability and utility of clinical guidelines for diagnostics and treatment of childhood aggression, according to academic experts’ perceptions (study 2.2.1); the diagnostics and treatment practices, and critical needs, according to mental health clinicians (study 2.2.2); and the availability of prevention and intervention programs for childhood aggression, according to academic experts’ perspectives (study 2.2.3). An additional study was compiled which reported on the awareness and perceptions of clinical guidelines, according to mental health clinicians, and overlapped this knowledge to that from experts’ reports on guidelines (study 2.2.4). Due to high variability in characterizing and diagnosing aggression in childhood across Europe, the umbrella concept of Severe Behavioral Problems (SBPs) was introduced. SBPs include frequent oppositional, aggressive and destructive behaviors, occurring across settings and impairing daily functioning.
2.2.1 Study Experts guidelines: ‘Awareness and perceptions of clinical guidelines for the diagnostics and treatment of severe behavioral problems in children across Europe: A qualitative survey with academic experts.
This study highlights the current successes and challenges perceived by experts around Europe associated with guidelines and documents for SBPs in children. Additionally, it establishes experts’ consensus for the need to develop official guidelines better tailored to clinical practice, creating an appropriate momentum for a transition towards European clinical guidelines for this population.
Some of the findings on guidelines availability have also been incorporated into a follow-up study and compared against mental health clinicians’ awareness of guidelines (see study 2.2.4. below).
2.2.2: Study Clinical practices: ‘Inventory of mental health clinicians' practices and critical needs for managing Severe Behaviour Problems (SBPs) in children across Europe’, according to mental health clinicians’ perspectives’
This project gathered feedback from mental health clinicians about diagnostic and treatment tools used and available in practice. Our recruitment reached 500 clinicians from 30 European countries. We received complete responses from 138 clinicians from 24 European countries, who reported through an online semi-structured questionnaire developed by the authors. Most clinicians had a medical training background (i.e. psychiatrists and psychiatrists in training; 53%), followed by psychotherapy and psychology. Clinicians perceived high severity of symptoms, comorbidity (e.g. ADHD, depression), poor daily functioning, and deficient family environment as risk factors indicating need for specialist treatment. Clinicians reported primarily using cognitive-behavior, pharmacotherapy and behavioral management programs as in clinical practice. Other specific programs used included the Incredible Years, Problem-Solving Skills Training, Anger Coping Program, VIG, and Triple-P. Clinicians also commented on critical needs in current practice. The main challenges include poor multi-agency collaboration, stigma and lack of awareness of SBPS, lack of resources, little support for teachers, and poor responsibility taking across systems. Clinicians’ main focus for future improvements included early detection and prevention (23%), multi-agency collaboration (21%) and increasing knowledge of SBPs (16%), although many clinicians had a pessimistic vision of a future (18%) with little funding and no prioritization of SBPs (12%). These findings highlight existing strategies and gaps in service provision for SBPs according to clinicians’ perspectives informing further action for improvement.
2.2.3. Study: Clinical practices: ‘Availability of prevention and intervention programs for severe behavioral problems in children across Europe, according to academic experts’ perspectives’
One interest of this study focused on in-depth investigation into experts’ opinions on accessibility of services and critical needs related to these programs. Numerous programs were mentioned, ranging from CBT-based interventions, Parent Training programs, or multi-systemic therapy to social services involvement and placement in residential care. Critical needs included improving implementation of these evidence-based programs by increasing their accessibility and increasing staff training, more evidence-based treatments, particularly behavioral-based parenting programs, more standardization of treatment, or wider dissemination. Differences across Europe were also identified with regards to how accessible services are for SBPs. Main challenges revolved around high demand over capacity. Overall, these findings provide a summary of available prevention and intervention programs across Europe, emphasizing existing gaps in service provision.
2.2.4 Study: Clinicians guidelines: ‘Perspectives on clinical guidelines for severe behavioural problems in children across Europe. A qualitative study with mental health clinicians’
We found that under half of the clinicians reported being unaware of any guidelines. Of these, 37% represented countries where guidelines were available according to experts. The remaining half of clinicians who were aware of guidelines on average reported being moderately familiar with their content, perceiving them as moderately useful and using them some of the time. Additionally, 60% clinicians agreed that SBPs guidelines need to be developed, as these would create a shared knowledge base for diagnosis, prevention and treatment. Guideline improvements included taking a multifactorial approach, creating specific case recommendations, encouraging multi-agency work. The need for additional funding, adaptation to local practices, and personalization to individual cases was also highlighted. These critical needs matched those highlighted by experts in study 2.2.1. Overall, the findings revealed that the modest familiarity with and use of guidelines amongst practitioners may highlight guidelines poor fitness-for-purpose, or, alternatively, an underlying confusion around the meaning and purpose of guidelines. Moving forward, efforts should be directed at disseminating clearer definitions of and instructions on using guidelines, and addressing current needs identified in this study.
Objective 2.3: Quantifying and identifying the long-term societal outcomes of childhood aggressive behavior.
Findings from objective 2.2 have substantially guided our activities in this objective.
2.3.1 Study Children with early-onset disruptive behavior: parental mental disorders predict poor psychosocial functioning in adolescence
The study provided novel evidence that parental MD puts 9-year-olds with DB at risk for negative outcomes in adolescence, both in terms of antisocial behavior and other outcomes, like school performance and truancy. Additionally, paternal MD was a better predictor than maternal MD, regardless of child DB at age 9, suggesting that fathers should be given increased attention in future research. Treatment-as-usual of children with DB could be augmented with additional screening and, if necessary, treatment of mental health problems in their parents.
2.3.2 Study The relation between the age at diagnosis of problem behaviors related to aggression and distal outcomes in Swedish children.
Severe childhood aggressive behaviors are known to predict negative outcomes later in life; however, little is known about the effect of when in childhood aggression problems are diagnosed. We analyzed data from the population-based Swedish Data Registries, which include data on all children formally diagnosed by the Swedish medical system with a wide range of aggression problems between ages 8 and 18 (N = 5,816) during the years 1987 to 2013, along with a matched control. Time-to-event analyses investigated whether the age at time of diagnosis affects later life outcomes while controlling for relevant confounders. Results show that for both boys and girls those with a later diagnosis had lower average incomes and a higher probability of having a criminal record. The effect on suicide attempts was not significant after correcting for multiple testing. Grade score was not significantly affected. The results warrant further research concerning advantage of earlier diagnoses and generalizability beyond the Swedish population.
2.3.3 Study Assessing Psychopathic Traits: Validation Study of The Child Problematic Traits Inventory and the Instrument of Reactive and Proactive Aggression in a Clinical Sample
Given the emphasis on empathy deficits, a clinical relevant focus related to empathy includes callous-unemotional (CU) traits. We ascertained whether CU measures are reliable and valid in clinic-referred youth and tested the psychometric properties of the Child Problematic Traits Inventory (CPTI) in clinic-referred children (ages 6-13 years). Teachers (N = 159) and parents (N = 173) completed the CPTI. Confirmatory factor analyses supported the CPTI’s 3-factor structure when teachers and parents rated the 28 CPTI items. Teacher- and parent-reported CPTI scores showed the expected relations with external correlates (e.g. conduct problems, and proactive aggression). Crucially, the validity of the CPTI scores was also supported across informants and across methods (i.e. regardless if a questionnaire or a diagnostic interview was used to measure external correlates). The CPTI holds promise as an assessment tool for assessing psychopathic traits in clinic-referred children. The parent- and teacher-reported Instrument of Reactive and Proactive Aggression (IRPA), offers another measure of severity of behavioral problems, and distinguishes between functions (reactive and proactive) and forms (physical/covert/verbal) of aggressive behavior. The IRPA will also be used to study subtypes of aggressive behavior within the clinical sample that has participated in the epigenetic and biomarker studies.
Objective 2.4:Develop long-term positive and negative trajectories from (early) childhood to late adulthood for subtypes and heterogeneity of aggression.
2.4.1. Study Can Mental Health Screening of Detained Male Juveniles Provide Evidence of Likelihood of Past and Future Violent Criminal Behavior?
In data from youth detention centers from the Netherlands we explored whether brief mental health screening tools are of value for alerting staff to a detained youth’s past offending or potential future violent offending. Boys (n = 1559; Mean age = 16.65) completed screening questionnaires that tap conduct problems, risk factors for aggression (e.g. anger-irritability), and mental health problems (e.g. depressed/anxiety). Official records were collected to index past and future violent offending. The small number of relations and their small effect sizes suggest little likelihood that screening for conduct problems, risk factors of aggression (e.g. anger-irritability) and mental health problems in boys who are detained in the Netherlands offers any potential for identifying past and future violent arrests.
2.4.2. Childhood aggression predicts externalizing but also internalizing adverse outcomes in emerging adulthood
In this longitudinal study, we examined the association between childhood aggression (based on parent-report; n = 18,649) with adverse outcomes in emerging adulthood identified in Swedish national registers, including rates of clinically diagnosed mental disorders, prescription of anti-depressant/sedative medications, suicide attempts, alcohol/substance use disorders, criminal convictions, failure to reach eligibility for high school, and social welfare support. We examined if childhood aggression predicted internalizing problems after adjusting for its overlap with externalizing problems and examined the genetic and environmental contributions to these associations using twin modeling. We found that childhood aggression predicted all outcomes in emerging adulthood. Childhood aggression was associated not only with a general problem factor and the independent externalizing factor, but also with the independent internalizing factor. The twin analyses showed that genetics contributed to the association between childhood aggression and all three outcome factors, whereas the shared environment seemed to be more important for the association between childhood aggression and the externalizing. The non-shared environment contributed marginally to all associations. These results indicate that childhood aggression is an important indicator of a host of adversities in emerging adulthood. Prevention and intervention of aggression at an early age may pay large dividends above and beyond aggressive behavior as it may also reduce anxiety, depression and several functional outcomes.
2.4.3. Harmonizing behavioral outcomes across studies, raters, and countries: an application to the genetic analysis of aggression in the ACTION Consortium
To include considerations of social and interaction effects requires large datasets with harmonized outcome measures. Here, we made use of a reference panel for phenotype data to harmonize multiple aggression measures in school-children to jointly analyze data from five large twin cohorts on 86,612 children (42,468 twin pairs) from five European twin cohorts. A bi-factor integration model in the integrative data analysis framework was developed to model aggression across studies while adjusting for rater, age, and sex. The harmonized aggression scores were analyzed to estimate contributions of genes, environment, and social interaction to aggression. The best-fitting model found a high level of overall heritability of aggression (~60%). Social interaction effects were significant in the opposite-sex twin pairs, and the model fit greatly improved when the social interaction effect of males on their female co-twin differed from the effect of females on their male co-twin. An aggressive female had a positive effect on male co-twin aggression, whereas more aggression in males had a negative influence on a female co-twin. The joint analysis with a large sample facilitated these new social interaction models for opposite-sex twins.
2.4.4. Effects of antipsychotics on future violence
Association Between Prescription of Major Psychotropic Medications and Violent Reoffending After Prison Release: This cohort study included all released prisoners in Sweden from July 1, 2005, to December 31, 2010, through linkage of population-based registers. Rates of violent reoffending during medicated periods were compared with rates during non-medicated periods in within-individual analyses. Follow-up ended December 31, 2013. Among released prisoners, rates of violent reoffending were lower during periods when individuals were dispensed antipsychotics, psychostimulants, and drugs for addictive disorders, compared with periods in which they were not dispensed these medications. Further research is needed to understand the causal nature of this association.
Second-generation antipsychotics and risk of violent crime in adolescents and young adults: Prescriptions of second-generation antipsychotics (SGAs) in young people have increased markedly during the last two decades, with a large proportion prescribed to those who exhibit disruptive behaviors and aggression. We investigated whether SGAs reduced the risk of violent crime in young people in a national cohort of individuals aged 15-24 years who have received at least one dispensed prescription of SGAs between 2006 and 2013 using data from Swedish national registers. Among young men received SGAs, rates of violent reoffending were lower during periods when individuals received SGAs compared with periods when they did not.
2.4.5. Ghrelin and aggressive behaviors
The relevance of ghrelin for human aggression per se as well as for aggression induced by alcohol was evaluated in a human genetic association study in young men (n=784) from the normal population assessed for anti-social behaviors. The study demonstrates that central ghrelin infusion, but not ghrelin administered systemically, increases aggression. Moreover aggressive behavior is decreased by pharmacological suppression of the growth hormone secretagogue receptor-1A (GHSR-1A). The genetic association study demonstrates that males carrying the Leu72Leu genotype of the pre-pro-ghrelin gene and displaying hazardous alcohol use are more aggressive when compared to the group carrying the Met-allele suggesting the identification of central ghrelin pathway as an important modulator in the onset of aggression.
Results for Objectives from our work package Genetic Epidemiology.
Objective 3.1: Select monozygotic twin pairs for sample collection for epigenetic and biomarkers studies
This objective was an essential step in the ACTION project. We collected buccal DNA and urine samples in a large group of twins concordant and discordant for aggressive behaviour. The buccal samples have been used to isolate DNA and to create methylation profiles. The methylation data have been cleaned and tested for association with aggression (Objective 3.5). The urine samples were collected for the biomarker research in WP5. Twin pairs from the large longitudinal database of the Netherlands Twin Register were selected. We targeted families with two control children (concordant control), two cases (concordant case) and families in which one of the twin pair is a control and the other a case (discordant pairs).
First-morning urine and buccal-cell DNA samples were collected across three phases of data collection: 1) the ‘technical pilot study’. This was a pilot which established feasibility and stability. Twins were unselected for phenotype information. First-morning urine and buccal-cell DNA was collected in twin pairs at two time points, with an interval of approximately two weeks. In total, 24 twin-children (12 twin pairs) were invited to participate. Permission for participation in the ‘technical pilot study’ was received from the parents of ten twin pairs (participation rate: 83.3%), 40% were girls. 2) the ‘biochemical study’. This was a large pilot to establish which biomarkers should be taken forward to the main study. We did collection of first-morning urine, buccal-cell DNA and assessed parent-reported aggression at the time of sample collection. Twin participants were invited as either aggression cases or controls, though complete twin pairs were always invited for participation. We invited 193 case-children, 395 control-children ( mean age of 9.8 years (SD = 1.6) 50.4% girls). The twins, of 293 families, belonged to 85 MZ twin pairs, 209 DZ twin pairs. Across all invitation we had a participation rate of 52% . We forwarded 88 participants to the next phase of the project, the ‘main study’. In total, the ‘biochemical study’ included 218 individuals of 6 to 11 years of age (mean age = 9.9; SD age = 1.6 51.2% females; 16 MZ, 93 DZ pairs). 3) the ‘main study’, collected data in MZ twin pairs concordant high concordant low on aggression and MZ twin pairs discordant on aggression. The mean age of 9.5 years (SD = 1.9) of the invited twins 49.6% were girls. The twins, from 1342 families, belonged to 1309 MZ twin pairs and 37 DZ twin pairs including 2 DOS twin pairs. Whilst BB3 aimed to include MZ twin pairs, DZ twin pairs were included for two reason: a) a selected MZ twin pair turned out to be a DZ twin pair after zygosity testing or b) the DZ twin pairs were siblings of a selected MZ twin pair and therefore also included in the study. Across all invitation batches we had a participation rate of 39.4% with 237 concordant low aggression twin pairs, 137 concordant high aggression twin pairs and 157 discordant twin pairs participating in the ‘main study’. In the ‘main study’ a total of 531 twin pairs agreed to participate, with the 44 twin pairs originally recruited as part of the ‘biochemical study’, and totals 575 twin pairs of 6 to 13 years of age (mean age = 9.5; SD age = 1.9 47.3% females) of which 255 were concordant low aggression twin pairs, 152 were concordant high aggression twin pairs and 168 were discordant twin pairs. There are 575 twin pairs (547 MZ and 30 DZ pairs).
The total response rate across all three phases of data collection for ACTION in the NTR was 42%. Urine and/or buccal-cell DNA samples (including 107 additional samples for epigenetic testing) are available for 1,495 twins (747 complete twin pairs), for whom 1,382 urine samples and 1,488 buccal-cell DNA samples were available. Twins have a mean age of 9.6 (SD age = 1.8) 47.5% of the participants were females (N = 710) and 82.6% belonged to a MZ twin pair (N = 1,234).
Objective 3.2: Unravel the genetic longitudinal architecture of aggression and estimate heritability as a function of age and gender
In order to fulfill objective 3.2 we took several approaches and used data of multiple ACTION cohorts. The main findings of the work under this objective are:
Genetic effects of childhood aggression are mostly stable, meaning that genes affecting childhood aggression are largely the same throughout childhood.
The heritability of aggression is similar in boys and girls, namely about 50-55% of the variability of the phenotype.
The indicators of aggression (i.e. questionnaire items) change in relevance throughout childhood (e.g. the item “threatens others” gains in importance). There are differences between boys and girls with respect to which items are the most relevant indicators of aggression. There is substantial differential heritability for the different indicators of aggression. For example, destruction of properties shows a very high genetic component during early childhood and fighting behaviors are more heritable in early adolescence.
A change in rater (mother vs. self-ratings) introduced complexity in the phenotype.
Measurement non-invariance across age can bias growth curve mean and variance estimates, and our quantification of this bias permits researchers to weigh the costs of using a sum score in longitudinal studies. Simulation results indicated that the genetic variance decomposition of growth factors is, however, not biased due to measurement non-invariance across age, provided the phenotype is measurement invariant across birth-order and zygosity in twins.
Across different informants, general and direct aggression at ages 12 and 14 predicted Antisocial personality disorder (ASPD) in a population-based sample.
Objective 3.3: Resolve patterns of comorbidity of aggression and emotional and behavioral problems as a function of shared genes and environment
In order to fulfill objective 3.3 we took several approaches and used data of multiple ACTION cohorts. The main findings of the work under this objective are:
Aggression co-occurs with the majority of other behavioral and social problems, from both externalizing and internalizing domains. We found neither rater differences nor differences across assessment instruments in co-occurrence patterns.
There were large similarities in co-occurrence patterns across European countries. Co-occurrence was also similar across age and sex, and if any change was observed, it indicated stronger correlations when children grew older.
For teacher ratings, correlations of aggressive behavior were high with other externalizing behaviors, and lower with internalizing problems. A negative association was seen with prosocial behavior.
Aggressive behavior at baseline was associated with higher BMI at follow-up. There was no evidence for reversed association. Aggressive behavior was prospectively associated with higher fat mass but not fat-free mass.
Over and above this list of findings of the ACTION project so far, there is also work in progress, including a study on the link between Insomnia and Aggression. Previous literature has shown a link between insomnia and all kinds of psychopathologies. We are conducting a study to test the hypothesis that a higher genetic risk for insomnia predicts aggressive behavior. With the latest summary statistics on an insomnia GWAS meta-analysis just released the analysists at each institution have just performed their prediction models and the meta-analysis is in progress.
Objective 3.4: Detect genomic regions of interest for aggression in children through genome-wide association (GWA) studies.
We took several approaches and used data of multiple ACTION cohorts and beyond. The main findings of the work under this objective are:
Overlap in item content across AGG measures ranged from absent (i.e. mutually exclusive) to moderate. Clinical concordance for the different AGG measures was very weak to weak. Rank correlations among continuous scores of AGG measures were moderate, suggesting stronger agreement when considering continuous scores compared to clinical cut-offs. Polychoric correlations revealed moderate to strong agreement between the different AGG measures. However, genetic correlations indicated substantial overlap in underlying genetic liability among the AGG measures.
A factor score meta-analysis can provide a gain in power over using sum score scores that were not directly comparable, provided the bi-factor model was adequately specified.
Our large meta-analysis of genetic associations with AGG from 29 cohorts was completed. It included 329,761 observations from 76,441 We report several suggestive loci on chromosomes 2, 3 and 13. The SNP-Heritability (h_SNP^2) of AGGoverall was estimated at 3,5%.
Rater-specific GWAMAs returned no genome-wide significant hits. Each rater-specific GWAMA, however, returned multiple suggestive loci. While some loci appear to be shared across multiple raters, none of the loci are common to all four rater-specific GWAMAs. Additionally, some loci are unique to their respective rater-specific GWAMA. The SNP-heritabilities range between 4% and 9.7% for different raters. Genetic correlations were moderate between AGGmother and AGGself (r_g=0.68; SE=0.10) and high between AGGmother and AGGteacher (r_g=0.83; SE=0.11). Interestingly, the genetic correlation between AGGself and AGGteacher was substantially lower (r_g=0.39; SE=0.12). The three rater-specific GWAMAs show largely the same genetic correlations with all outcomes.
Genetic correlations of childhood aggression with other traits give a meaningful picture of the genetic liabilities that are shared with childhood aggression: higher levels of AGG in children are associated with a much increased genetic risk of ADHD, and of developing a wide range of psychopathologies in adulthood. Children with high AGG have larger risks of loneliness, smoking, poor self-rated health, obesity and sleep problems.
Objective 3.5: Identify differentially methylated regions on the human genome related to differences in aggression using a genome-wide approach.
DNA for this objective was collected (see 3.1) in the same group of twin pairs and clinical cases as for the metabolomics projects. The samples have been measured on the Illumina EPIC 850K array. QC of the data has been performed with the pipeline developed by us in collaboration with the Dutch BBMRI-BIOS consortium (https://www.bbmri.nl/acquisition-use-analyze/bios/). Pipeline development and adjustment was required because Illumina stopped the production of the 450K array on the launch of their new Epic (850K) during the ACTION project. To adapt data processing and analysis pipelines for the EPIC array, we performed a study on DNA samples (from existing NTR DNA collections for twins with buccal swaps and aggression phenotyping) measured on the Illumina EPIC array. The samples were from 53 monozygotic twin pairs (age 0-10), scoring high or low on aggressive / rule breaking behavior. For one twin pair, the same DNA sample was run twice on to examine test-retest reliability of the array. We performed the first genome-wide analysis of monozygotic (MZ) twin correlations and mQTLs on data obtained with the Illumina MethylationEPIC BeadChip (EPIC array) and compared its performance to the Illumina HumanMethylation450 BeadChip (HM450 array) for buccal-derived DNA. We conclude that the contribution of familial factors to individual differences in DNA methylation and the effect of mQTLs are larger for novel EPIC probes, especially those within regulatory elements connected to active regions specific to the investigated tissue.
Epigenome-wide association study for Aggression: While the new DNA samples were being collected and measured, we performed the first large EWAS meta-analysis (EWAMA) on existing data of aggressive behavior and genome-wide DNA methylation assessed with the Illumina 450k array (mainly) or the EPIC array. We contacted all possible collaborators around the world for the availability of Methylation and aggression data, which resulted in the inclusion of 21 cohorts with data for 14,434 peripheral blood samples (age at blood sampling 7 to 70) and 2425 cord blood samples in the EWAMA. For follow-up analysis we collaborated with the other consortia that received funding in the FP7 Work program HEALTH.2013.2.2.1-3: Paediatric conduct disorders characterised by aggressive traits and/or social impairment: from preclinical research to treatment. Data in two clinical cohorts were available in NeuroIMAGE and FemNAT-CD and brain methylation data from rodent models of aggression from the MATRIX consortium. For the discovery analyses, the age at phenotypic assessment ranges from 4 to 68 and multiple instruments were used to assess aggressive behavior or conduct problems.
EWAMAs were performed for whole blood, cord blood, and for all data combined. These analyses identified 48 top CpGs for which we performed enrichment analyses. The top ten most strongly enriched traits indicate that CpGs associated with aggressive behavior overlap with CpGs previously associated with smoking, maternal smoking, and other chemical exposures (e.g. perinatal exposure to polychlorinated biphenyls (PCBs) and polychlorinated dibenzofurans (PCDFs)). Further overlap includes CpGs associated with alcohol use, cognitive function, educational attainment, ageing, and metabolic traits. Next steps involve integration of the EWAMA findings with the novel EPIC methylation data collected in D3.1. The genome-wide summary statistics from this project will be made available to the scientific community.
Epigenome-wide association study for ADHD symptoms: On existing data we also performed the first epigenome-wide association study meta-analysis of adult attention-deficit/hyperactivity disorder symptoms, based on peripheral blood DNA methylation (Infinium HumanMethylation450K array) in three population-based adult cohorts: the Netherlands Twin Register (N = 2258, mean age 37 years), Dunedin Multidisciplinary Health and Development Study (N = 800, age 38 years), and Environmental Risk Longitudinal Twin Study (N = 1631, age 18 years).
One epigenome-wide significant differentially methylated position was detected in the Dunedin study, but meta-analysis did not detect differentially methylated positions that were robustly associated across cohorts. In region-based analyses, six significant differentially methylation regions (DMRs) were identified in the Netherlands Twin Register, 19 in the Dunedin study, and none in the Environmental Risk Longitudinal Twin Study. Of these DMRs, 92% were associated with methylation quantitative trait loci, and 68% showed moderate to large blood-brain correlations for DNA methylation levels. DMRs included six non-overlapping DMRs (three in the Netherlands Twin Register, three in the Dunedin study) in the major histocompatibility complex, which were associated with expression of genes in the major histocompatibility complex, including C4A and C4B, previously implicated in schizophrenia.
Results for Objectives from our work package Gene-Environment interplay
Objective 4.1: Identify and replicate ‘pure’ environmental risks that have their effects on the development of aggression independent of genetics
ACTION twin analyses have contributed to the research literature demonstrating that most of the environmental variance in aggression and other behavior problems is not shared by children growing up in the same family. We have exploited the powerful identical (MZ) twin differences design to identify pure environmental factors contributing to aggression. We have now completed 10 extensive interviews of MZ twin pairs who are discordant at 1.25 standard deviations for measures of aggression assessed at age 16. Interviewers explored the family, school and friend-wide factors that each member of a twin pair separately attributed to their discordant aggression scores. Crucially, each member of a twin pair reported on their own aggressive behavior but also on their co-twin’s aggressive behavior, which allowed for corroboration of environmental contributors across raters. Analyses of our in-depth interview data suggest adolescents’ aggressive behaviour is likely to be influenced by several risk factors spanning multiple areas of their lives. As a result, we harnessed the predictive power of multiple risk factors, namely: environmental niches (school, home and neighbourhood characteristics); (2) growth and health (BMI, pubertal development, physical and mental health); and (3) psychological resilience (psychological adjustment, wellbeing and motivation) to create a multidimensional risk score (MRS) for 5,635 twin pair members of the Twins Early Development Study (TEDS). We conducted twin model-fitting analyses of these data.
MRS predicted a similar portion of variance in concurrent aggression at both ages 12 and 16 (explained variance R2 = .10 and R2 = .11, respectively). The aetiology of the association between MRS and aggression differed between early and late adolescence. While shared environment explained most of the covariance at 12 (51%), genetic factors were the main source of covariance at 16 (50%). Although non-shared environment was the main factor in the aetiology of aggression at 16 (e2 = 72%), it was only minimally implicated in the overlap between MRS and aggression. Our interviews with 10 MZ pairs highly discordant for aggression suggest that non-shared environments are largely idiosyncratic, such as differences in romantic relationships, differences in the use of oral contraceptive or differential responses to parental separation. Thus, risk factors of aggression span multiple domains of adolescents’ lives. Although MRS predicted a significant portion of variance, the effect is modest, emphasising the need for future research to identify additional risk factors. The observed lack of non-shared environmental overlap with aggression suggests that non-shared environment is idiosyncratic. This in-depth analysis of the individual and collective role of socio-ecological, physical and psychological characteristics to variation in aggression is likely to provide essential new knowledge on key targets for environmental interventions aimed at reducing the personal and societal burden associated with aggressive behaviour.
Objective 4.2: Identify and replicate environmental effects on the development of aggression that depend on genetic propensities (GxE)
GxE using Twin Method: In order to address GxE we first we examined phenotypic correlations between environmental measures separately at ages 7, 9, 12 and 16 and verbal and physical aggression at age 21. We then compacted the analyses in a multiple regression approach that examined the combined prediction from all childhood environmental measures at each childhood age on verbal and physical aggression at age 21. At each age and across ages, we created a novel ‘poly-environmental’ risk score for age-21 aggression. To test the joint prediction of childhood environmental measures on age-21 aggression, we used a leave-one-out multiple regression approach. That is, regression weights from 90% of the sample were used to create a poly-environmental risk score for the other 10% so that scores were derived from an independent sample. This poly-environmental score was calculated at ages 7, 9, 12 and 16. We then explored whether the heritability of aggression differs at different levels of the poly-environmental risk score using a continuous GxE analysis (Purcell, 2002). For example, does high poly-environmental risk for aggression lower heritability by overwhelming genetic propensities?
Results indicated that Poly-environmental scores at each age correlated modestly with aggression at age 21 (average correlation = 0.2). The correlations were of similar magnitude for physical and verbal aggression. These are ‘main effects’ of the childhood environment on adult aggression.
We examined the extent to which the poly-environmental risk scores moderated the manifestation of genetic and environmental effects on aggressive behaviour during childhood and adolescence. The results showed significant GxE interaction effects during childhood. At age 7, the twin heritability of aggression was observed to increase at higher levels of exposure to poly-environmental risks factors, going from ~.20 at lower levels of environmental risk (-2SD) to ~.60 at high levels of exposure to environmental risk (+2 SD). A similar pattern of results was observed at age 9, where the heritability of aggression was ~.30 at low levels of exposure to environmental risk and ~.60 in twins exposed to higher levels of poly-environmental risk. However, these interaction effects did not replicate in adolescence, when the twins were 12 and 16 years old. At these ages, twin heritability estimates remained consistent (between .3 and .4) at different levels of exposure to poly-environmental risk. We conclude that childhood environments predict adult aggression but the evidence for moderation effects of genes on adult aggression is limited. We detected significant GxE for aggression at ages 7 and 9, but this did not replicate at ages 12 and 16 in the TEDS sample. Further replication in other ACTION samples is planned to test for the specificity of detected GxE to early childhood years.
GxE using Polygenic Score method: Polygenic scores derived from major GWA studies of adult psychiatric disorders were used to create a polygenic score for ‘p’, a general factor of psychopathology, for each individual in TEDS. We then assessed GxE interaction between the polygenic score for p and the environmental measures in terms of their prediction of aggression. Unlike twin analyses, we were able to investigate GxE for all of our environmental measures regardless of whether they were child-specific or family-wide. We tested poly-environmental scores at ages 7, 9, 12 and 16 (see Study 1) in our analyses of GxE. We did not find significant GxE effects. That is, the prediction of aggression from the environmental measures was not conditional on the polygenic score for p. Thus, for the measures used in our analyses, we can conclude that GxE for aggression does not emerge in systematic analyses using polygenic scores as well as using the twin method. However, novel environmental factors, most notably therapeutic interventions, could show stronger interactions with genetics. The key advance needed for polygenic score research on aggression is to create more powerful polygenic scores, but this advance depends on increasing the SNP heritability of aggression, which is low (<10%).
GxE using novel methods: Because of difficulties in identifying GxE, we developed new approaches using atheoretical machine-learning approaches to detect interactions across diverse environmental measures, polygenic scores, and phenotypic measures. First, we perform Elastic Net regularization, including covariates such as age, sex and 10 genetic principal components, independent variables (polygenic scores and environmental predictors), and the bivariate interaction terms for covariate*predictor and predictor*predictor. Second, we fit a random forest regression to the data, including covariates and predictors (PGS and measured environments), leveraging the ability of regression trees to automatically detect interaction patterns in the data, as well as non-linear effects. For each phenotype, all models are trained in 80% of the data and tested for performance (with respect to prediction accuracy) in the remaining 20%. Analyses include proximal and distal environmental measures such as life events, home environment, and socio-economic status, along with polygenic scores for major psychiatric disorders (e.g. schizophrenia), cognitive traits (e.g. intelligence) and anthropometric traits (e.g. body mass index). We conducted these analyses separately for each trait. Prediction models were constructed for behaviour problems (including aggression), psychiatric disorders and anthropometric domains. For every phenotypic outcome, we fit linear and decision-tree based models integrating environmental and polygenic predictors, assessing their predictive accuracy while testing for (multivariate) interactions. Preliminary results from our prediction models indicate that this new approach yields many suggestive GxE but of very small effect size. When the ACTION+ GWAMA results are available, we will include a polygenic score derived from the summary statistics in these analyses. Follow up linear models (such as OLS regression) for all detected interactions are currently being conducted in replication cohorts in ACTION.
Objective 4.3: Identify and replicate genetically driven experiences associated with the development of aggression in which children select, modify or create environments correlated with their genetic propensities towards aggression (rGE)
rGE using Twin Method: We systematically explored gene-environment correlation (rGE) using bivariate twin analyses for a wide range of environmental measures as they predict self-reported verbal and physical aggression at age 21. We found that 90% of the correlation between the poly-environmental risk score and aggression could be accounted for by genetic factors. Similar results were found in bivariate twin analyses of those individual childhood environmental measures that predicted adult aggression. These findings strongly implicate rGE in the prediction of adult aggression from childhood environments.
rGE using Polygenic Score Method: Twin analyses of rGE are limited to environmental measures that differ within families (i.e. child-specific measures). Here we used a polygenic score for ‘p’ and a general factor of psychopathology in analyses of rGE. We first asked if the polygenic score for ‘p’ was, as expected, associated with our measure of aggression at age 21. We then tested for an association between the polygenic score for ‘p’ and the poly-environmental risk score for aggression. We found that the polygenic score for ‘p’ was significantly associated with aggression but predicted less than 0.5% of the variance in aggression, which is comparable to the variance explained for other measures of behaviour problems. A correlation of .05 was found for the association between the polygenic score for ‘p’ and the poly-environmental risk score for aggression, indicating rGE. This polygenic score for ‘p’ correlates as highly with the poly-environmental risk score for aggression as it correlates with measures of behaviour problems including aggression. A major limitation of these analyses is that the polygenic score is only weakly associated with aggression, which makes it difficult statistically to show that the association between childhood environmental risk and adult aggression. The key advance needed for polygenic score research on aggression is to create more powerful polygenic scores, but this advance depends on increasing the SNP heritability of aggression, which is low (<10%). Nonetheless, for the measures used in our analyses, we can conclude that rGE for aggression is widespread in systematic analyses using polygenic scores as well as using the twin method.
rGE using DZ twins: Polygenic score prediction among unrelated individuals may include contributions from both direct genetic effects and also indirect effects due to passive GE, as well as population stratification and assortative mating. Here we used the within-family association between sibling differences in polygenic scores and sibling trait differences, which is free of such indirect effects of shared environment, to test for rGE. The most relevant finding is that evidence was found for passive rGE for cognitive traits (intelligence and educational achievement) but not for any other traits such as personality or behaviour problems, including physical and verbal aggression. The implication for aggression is that passive rGE is unlikely to be an important factor in the origins of individual differences in aggression.
Objective 4.4: Apply GCTA methods, GWA genotype data, and polygenic predictors of aggression phenotypes to confirm the results of twin-family analysis of GxE and rGE and to identify some of the genes responsible for GxE and rGE
By leveraging a SNP-derived genetic relationship matrix (GRM) we compared genetic similarities between unrelated individuals to similarities on verbal and physical aggression to infer the contributions of SNPs to the variation in these traits. SNP heritability analyses for the BAQ self-report measure at age 21 yielded an exciting result: the SNP heritability for verbal aggression was 32%, much higher than typical SNP heritability estimates of about 10%. However, when we tested this result in the Queensland sample, which included the same BAQ measure, the result did not replicate, demonstrating the value of ACTION for immediate replication research. SNP heritability for aggression is about 10%, typical of other behaviour problems, which is important because SNP heritability is the ceiling for GWA studies.
Objective 4.5: Extend analysis of epigenetics (WP3) to investigate GxE and rGE
DNA methylation data from the ACTION EWA meta-analysis (EWAMA) of aggression were used to create polygenic scores for TEDS twins by weighting the corresponding top loci that emerged from the EWAMA (i.e. SNP-CpG associations) by their effect sizes (z-scores), using the –score function in Plink (epitot). We also created two additional scores, one based on the CpGs positively associated with aggression (epipos) and one with the ones negatively associated with aggression in the EWAS (epineg). Our SNP polygenic score for the top methylation sites was not significantly associated with verbal, physical or total aggression at age 21. This polygenic score was also not correlated with our poly-environmental measure (rGE), nor did it mediate the modest correlation between the poly-environmental measure and aggression. The EpiGPS scores were not associated with aggression, aggression subtypes, polygenic p, or our multi-environmental scores. Ongoing analyses are investigating GxE and rGE using epigenetic scores from the ACTION EWA of aggression.
Results for Objectives from our Biomarker work package.
Objective 5.1: to identify biomarker profiles to diagnose aggression
A review of the literature was performed to identify known biochemical biomarkers for human aggressive behavior . This identified four classes of biochemical biomarkers potentially involved in human aggressive behavior: 1) inflammation markers (e.g. interleukin 6, C-Reactive Protein); 2) neurotransmitters (e.g. 5-HIAA, HVA, glutamate); 3) lipids (e.g. HDL, LDL); and 4) hormones (e.g. cortisol, testosterone). Methylation markers have also been suggested to be associated with aggressive behavior. Furthermore, Dekkers et al. (Genome Biol., 2016) showed that methylation and blood lipid levels are not independent and that differential methylation can be a consequence of variation in blood lipid levels. Therefore, we hypothesized that the methylation levels of such loci in blood can inform us if aggression is associated with long-term exposure to lipid level. If this is the case, we expected to find that loci where methylation levels were influenced by lipid levels show differential methylation in aggressive individuals. Such loci might complement classic lipid level measures as a biomarker for lipid-related disturbances in aggression.
As a first step, we examined the association of lipid levels and related biomarkers with aggression in existing data from a large adult population cohort (N = 5,588), including 31 monozygotic (MZ) twin pairs who were discordant for aggression, as well as 12 extremely discordant MZ pairs. Biomarkers were not significantly associated with aggression in the population cohort. In the discordant MZ pairs we identified significant within-pair differences for glucose and marginally significant differences for lipids and cytokines, with the more aggressive twin showing lower levels of glucose and LDL and higher levels of fibrinogen, C-reactive protein and interleukin-6. The analysis of epigenetic data in the MZ pairs discordant for aggression did not show enrichment for lipid cytosine guanine dinucleotides (CpGs) and we observed no enrichment of lipid CpGs in an epigenome-wide association study of aggression in the population cohort.
In the Biochemical Study we analyzed urine samples of 222 children (mean age = 9.9; SD age = 1.6) selected for the absence or presence of aggression problems. We reclassified the case-control status of twins based on their mother-rated aggression scores at the time of urine and DNA collection. This classification identified 110 cases, 95 controls (17 children without new status). To identify potential biomarkers for childhood aggression we developed LASSO GLM models for the organic acids and amines metabolomics platforms. These models identified no biomarkers based on the amines platform and selected 4 metabolites (α-hydroxybutyrate, succinic acid, aspartic acid and uracil) based on the organic acids’ platform. The predictive accuracy (AUC) of the organic acids model is relatively was 67% and including these metabolites in a model with sex and age did improve the predictive accuracy compared to a model including only sex and age. Univariate follow-up analysis including age, vitamin use and batch effect as covariates found a significant association with aggression for α-hydroxybutyric acid (β = 0.31 p = 0.04).
In the Main Study we analyzed urine samples from 1,348 twins (mean age = 9.6; SD age = 1.8) selected for their concordance or discordance for aggression problems. In addition, the Main Study included 186 children of 6 to 13 years of age (mean age = 9.6; SD age = 1.8) referred to the clinic. To identify biomarkers for childhood aggression in the Main Study we used a three-step approach: 1) between-family analyses; 2) within-family analyses and 3) replication analyses.
For the median-normalized metabolites, prior to correction for multiple-testing, the between-family analyses showed significant negative associations of L-methionine-sulfoxide (estimate = -0.003; Robust SE = 0.002; p = 0.03) and glyceric acid (estimate = -0.001; Robust SE = 0.001; p = 0.04) with childhood aggression. We observed no significant differences in metabolite concentrations between aggressive and non-aggressive twins for the top 25% of the amines (16 metabolites) and organic acids (5 metabolites) in the within-family analyses. Congruent directions of effect in the between- and within-family analyses were observed for 2 of the 5 top 25% organic acids and 7 of the 16 top 25% amines. Repeating the within-family analyses with only discordant MZ twins improves the congruency of the direction of effect among the analyses to 16 of the 21 top 25% metabolites. Replication of the top 5 metabolites across both platforms (3-methyladipic acid, cystathionine, cysteine, glyceric acid and homocysteine), in 186 clinical cases and 93 twin pairs concordant low for childhood aggression, confirmed that amines and organic acids were not significantly associated with childhood aggression in the ACTION project.
For the density-normalized metabolites and biomarkers, after correction for multiple testing the between-family analyses showed associations of childhood aggression with O-phosphoserine (estimate = 0.33; Robust SE = 0.09; p = 0.02) Gamma-L-glutamyl-L-alanine (estimate = 0.31; Robust SE = 0.09; p = 0.02) and Gamma-Glutamylglutamine (estimate = -0.16; Robust SE = 0.05; p = 0.03). The concentrations for these 3 metabolites were not significantly different between aggressive and non-aggressive twins. Replication analyses were performed for the top 5 metabolites most associated metabolites with consistent direction of effect in the between- and within-family analyses in 186 clinical cases and 93 twin pairs concordant low for childhood aggression. We observed that norepinephrine (mean cases = 0.37; mean controls = -0.15; p = 3.03x10-5) saccharopine (mean cases = 0.20; mean controls = -0.08; p = 0.01) ethanolamine (mean cases = 0.21; mean controls = -0.08; p = 0.01) and gamma-L-glutamyl-L-alanine (mean cases = -0.01; mean controls = -0.23; p = 0.03) different significantly among cases and controls after correction for multiple testing, with congruent direction of effect across all 3 analyses.
In conclusion, based on the existing literature we identified 4 classes of potential biomarkers for (childhood) aggression: 1) lipids, 2) inflammation markers, 3) neurotransmitters and 4) hormones. With the work performed in ACTION WP5 we find further suggestive evidence for the role of inflammation markers, lipids and amino acids involved in neurotransmitter pathways in aggression. In children we identified metabolites of the cysteine metabolism pathway, involved in oxidative stress, as among the most promising aggression biomarkers and in adults we found cytokines to be marginally significantly related to aggression differences in extremely discordant MZ pairs. LDL cholesterol levels were lower (marginally significantly) for the MZ twins scoring high on aggression as compared to their co-twin scoring low on aggression. Finally, the top 25% most associated metabolites in ACTION’s Main Study included multiple (essential) amino acids which are involved in neurotransmitter pathways, or may act as neurotransmitters themselves, such as the metabolites included in the cysteine metabolism pathway.
Objective 5.2: to identify biomarkers to differentiate between sub-groups of aggression
The most common subtypes of aggression studied to date are verbal vs. physical aggression, impulsive vs. premeditated aggression or reactive vs. proactive aggression. Our literature review suggested that lipids, neurotransmitters and hormones are associated with subtypes of aggression, however, the literature does not suggest any clear pattern of biochemical biomarkers associated with subtypes of aggression. For example, decreased cortisol levels have been associated with both reactive and proactive aggression, while increased cortisol levels have also been associated with reactive aggression.
In order to add the additional steroid targets suggested by literature to the target list measured in the biochemical study, development on a novel steroid platform measuring both conjugated and unconjugated forms was undertaken. Steroids in this platform have a wide range of polarities and extensive development work was necessary. Getting the proper standards for the targeted LC-MS/MS was difficult as many steroids are controlled compounds and took a longer time than originally expected, but the method is now being validated and the samples are measured. The target list that has been finally chosen is based on input from various discussions within ACTION, with the scientific advisors of ACTION, and other experts in the field, and the target list is based on availability of the steroids as reference compound. The developed platform will now be used for samples collected as part of the ACTION project and will also be available for future studies.
To identify subgroups for aggression we defined subgroups based on our work of the comorbidly with other psychiatric traits, specifically symptoms of attention-deficit/hyperactivity disorder (ADHD). In 1,329 twins we investigated the role of biomarkers and metabolites on aggression comorbid with ADHD-symptoms. This comorbidity phenotype was classified as follows: “low aggression and low ADHD-symptoms”, 2) “high aggression and high ADHD-symptoms”, 3) “low aggression and high ADHD-symptoms”, and 4) “high aggression and low ADHD-symptoms”. GEE analysis of the aggression comorbidity phenotype showed that 14 median-normalized metabolites levels associated with aggression and/or ADHD-symptoms as compared with controls (children with low aggression and ADHD-symptoms). Higher levels of 3-methylhistidine, adipic acid, anserine, glyceric acid and histamine, as well as lower levels of cystathionine and cysteine were observed for children with high aggression and high ADHD-symptoms. Lower levels of 3-hydroxybutyric acid, 3-methyladipic acid and fumaric acid were observed for children with high aggression but low ADHD-symptoms. Low aggression and high ADHD-symptoms were associated with increased levels of L-isoleucine and L-threonine. Results for L-leucine and L-methionine-sulfoxide were inconclusive as higher metabolite levels were observed for multiple subgroups.
GEE (generalized estimating equation) analyses on the density-normalized metabolites and biomarkers indicated that lower taurine, gamma-glutamylglutamine, epinephrine and 3-hydroxyisobutyric acid levels and higher o-phosphoserine levels were significantly associated with high aggression and low ADHD-symptoms. Lower levels of citric acid and DL-3-aminoisobutyric acid were observed for children with low aggression and high ADHD symptoms. However, results were inconclusive for cysteine, fumaric acid, glycylglycine, homocysteine, malic acid, glutathione and cystathionine as these were associated with multiple subgroups.
In conclusion, with the work performed in ACTION WP5 we find further suggestive evidence for the role of amino acids involved in neurotransmitter pathways in aggression comorbidity. We found that levels of the essential amino acids L-isoleucine and L-threonine were associated with ADHD-symptoms, but not with aggression and observed that several metabolites, including metabolites of the cysteine metabolism pathway, were associated high aggression and ADHD-symptoms. Furthermore, the suggestive role of the cysteine metabolism pathway suggests that inflammation (oxidative stress) is also involved in aggression and ADHD-symptoms.
Objective 5.3: identify treatment options for (different sub-groups) of aggression
The amino acids isoleucine, leucine, methionine, tyrosine and valine had lower levels in children with high aggression as compared with low aggression (Deliverable 5.1). All of these amino acids are (conditionally) essential. Supplementation of these (conditionally) essential amino acids to increase their levels might help to reduce childhood aggression. Specifically, supplementation of methionine may be beneficial to alleviate aggression in children. Methionine is a precursor in the cysteine metabolism pathway. Methionine supplementation therefore might increase homocysteine levels which in turn might rebalance the disturbed cysteine metabolism pathway. With the balance in the cysteine metabolism pathway restored the potential disrupted antioxidant glutathione levels may be restored. Therefore, methionine supplementation might restore the balance between antioxidant and reactive oxygen species thus alleviating oxidative stress. ‘Subtype’ analyses show that restoring the balance to the cysteine metabolism pathway may be especially beneficial for those children with comorbid high aggression and ADHD-symptoms.
Results for Objectives from our work package Novel treatment and prevention strategies.
Objective 6.1: Development of evidence-based model on effectiveness of existing preventative and treatment strategies for paediatric aggression through meta-analyses (1. Prevention, 2. Intervention)
To extend existing knowledge on prevention and intervention of childhood aggression, we did a literature synthesis of 72 systematic reviews and meta-analyses on treatment effectiveness for childhood aggression. For universal and selective prevention, treatment effectiveness was absent or weak; for indicated prevention and intervention, treatment effectiveness was mostly small or medium. Furthermore, most moderators of treatment effectiveness had no effect in the majority of studies (i.e. child age, child gender, implementation to individuals or groups, different treatment programs, person implementing the treatment, and session related factors or treatment intensity) or mixed effects (i.e. socioeconomic status, type of treatment, informant, research quality). The only two significant moderators comprised of positive effects of higher pre-treatment levels of aggression and parental involvement. Findings suggested that future research distinguishing between targets of the treatment (i.e. factors associated with childhood aggression; universal and selective prevention vs. present aggressive behaviors; indicated prevention and intervention) would be promising. In addition, more knowledge on parental influences is needed, including parental psychopathology, parenting style, and heritability .
Objective 6.2: Integration of insights from WP2-WP5 in a comprehensive theoretical framework to inform novel prevention and intervention efforts .
We collaborated on a project that combined data from CATSS and NTR (total sample size was 62,227 children) to predict childhood aggression using a novel methodological approach. The outcome measure was the phenotypically harmonized overt/physical aggression factor score created within Action. The large set of predictor variables encompassed demographics, prenatal characteristics, physical development, parental education level, life events, and mother-reported behavioral symptoms. To avoid capitalization of chance, data were partitioned in four parts for the different analysis steps. These included 1) exploratory data analysis and tuning meta-parameters for the data mining, 2) fitting increasingly complex data mining models to assess which predictors had which types of effects (i.e. linear, nonlinear, interaction), 3) assessment of model performance and importance of the predictor variables, and 4) fitting a confirmatory prediction model of aggression that integrated the results of the data mining analyses. The resulting multi-group model accounted for interactions with sex and cohort and confirmed linear main effects of variables measuring behavioral symptoms (e.g. related to non-physical aggression, attention-deficit/hyperactivity disorder, and conduct disorder), maternal smoking during pregnancy, parenting, and proportion of life events. The most important predictors comprised behavioral symptoms such as arguing, being easily distracted, and hyperactivity. Findings were in line with previous research, yet weaker, likely due to the simultaneous analyses of many predictors. These easily observable predictive behaviors may act as targets for early detection and prevention of childhood aggres. These findings confirmed that childhood aggression cannot be viewed in isolation from other emotional and behavioral problems.
Objective 6.3: Development of risk assessment charts together with guidelines to improve clinical decision making
Development of childhood aggression is shaped by interactions between individuals and multiple contexts; a non-exhaustive overview can be found in the Table below. These contexts range from the proximal (e.g. family) to the more distal contexts (e.g. policies for child and adolescent mental health). To examine the extent to which national-level policies for child and adolescent mental health (CAMH) are associated with individual-level aggressive behavior, we analyzed data from 172,829 adolescents (aged 11 to 15 years) in 30 European countries from the Health Behavior in School-aged Children study. Information on national-level policies for CAMH was gathered from renowned statistical institutes and included availability of epidemiological data, the number of CAMH facilities, investment in family benefits, and investment in education. Multi-level analyses revealed lower adolescent aggressive behavior in countries with more CAMH policies. This association held when taking into account other national-level indicators, including adult violence, adult well-being, and income inequality. We translated these findings and previous literature into guidelines to improve clinical decision making.
An established risk factor for childhood aggression is lower socioeconomic status (SES). There is, furthermore, some evidence that SES has a moderating effect on the causes of individual differences in childhood aggression, with lower heritability estimates in low SES groups and a stronger contribution of environmental factors. In collaboration with researchers from WP3 and WP4, we conducted a study on the moderating effect of socioeconomic status (SES) on the genetic architecture of childhood aggression among 7-year-old twins in the Netherlands and the United Kingdom. We found that, compared to high SES, for low and medium SES the standardized contribution of the shared environment to individual differences in childhood aggression was higher. This finding suggests that aggressive children from low and medium SES may benefit more from interventions targeting their surroundings than aggressive children from high SES.
Our review on the effectiveness of interventions highlighted the importance of parent and family factors. Family factors e.g. inter-parental violence, family conflict and harsh parenting, contribute to childhood aggression by diminishing children’s self-control. At the same time, children with low self-control evoke more harsh and negative parenting behavior and family conflict. In addition, following up on our review, we contributed to a study (WP2) that investigated if parental mental disorders (MD) increase risk of future maladjustment among children who early-onset disruptive behavior (DB). In a large, longitudinal study, parents of 9-year-old children reported on child DB, while a patient registry was used to determine parental MD. At follow-ups at 15 (N = 6319) and 18 years (N = 3068) information about various problems was collected via registries, parent- and, self-reports. Results revealed that parental MD puts 9-year-olds with DB at risk for negative outcomes in adolescence. Additionally, paternal MD was a better predictor than maternal MD, regardless of child DB at age 9, suggesting that fathers should be given increased attention in future research. The findings also suggest that the effectiveness of prevention and treatment efforts for children with DB can be enhanced with additional screening and, if necessary, treatment of parental MD.
Objective 6.4: Assessment of the empirical theoretical framework and risk assessment charts with respect to ethical, legal and social implications (ELSI)
To address the ethical, legal, and social implications (ELSI) of the research conducted within ACTION, we contacted all partners and reviewed all publications and ACTION manuscripts. The outcomes are discussed in Deliverable 6.3. In brief, ethical implications of ACTION comprised of the concept of childhood aggression (i.e. multidimensional disorder, highly co-morbid with other emotional and behavioural problems likely due to a shared genetic liability, greater agreement among assessments of aggression using a continuous approach), patients’ understanding (i.e. when genetics and biomarkers are implemented in treatment all stakeholders need to understand what it means), and the potential for stigmatization and the harmful effects of labelling (i.e. prediction models need to reach high precision and accuracy before the benefits outweigh costs of misclassification). Legal implications concerned redefinition of existing regulatory policies (i.e. more research is needed to understand the association between policies for child and adolescent mental health and aggression), professional ethical guidelines (i.e. academic experts expressed a need for national-level and Europe-level guidelines for severe behaviour problems, with reinforcement for implementation, tailored to daily practice), and inclusion of biomarkers in clinical decision making (note that biomarkers with such predictive value have not yet been identified in ACTION research). Social implications comprised stigmatization and labelling effects (i.e. subpopulation identified for heightened risk of aggression is children whose parents have a psychiatric disorder, also an important factor for treatment effectiveness), and disparity, diversity, and equity (i.e. results may not generalize to minority groups, etiology of aggression differs among SES strata so may also differ for minority groups). Although the research within ACTION does not yet have implications that can directly be translated into ELSI, the built collaborations and the wealth of data do provide sufficient tools to work towards improvements in prevention and intervention efforts and identification of actionable ELSI in the future.
RISK FACTORS ASSOCIATED WITH AGGRESSION IN CHILDHOOD AND ADOLESCENCE
Individual factors
Biological characteristics Psychological and behavioral characteristics Demographics
Biomarkers including metabolomic profiles
Genetics (heritability, GxE)
Genetics (GWAS, epigenetics)
Complications/risks during pregnancy (e.g. maternal smoking; birth weight; maternal age)
Resting heart rate
Puberty
Body size
Attention problems, lack of behavioral control (impulsiveness, low executive functioning).
Arguing, being easily distracted, hyperactivity
Low IQ
Low educational performance
Low school commitment
Drug use (esp. at an early age)
Callous Unemotional traits Sex
Age
Social factors
Family influences Parental conflict Peer influence
Parenting:
Parental warmth
Control
Harsh discipline, spanking
Monitoring, supervision
Neglect Divorce
Domestic violence Having delinquent friends or siblings
Drug use
Family characteristics
Demographics and structure Parental characteristics
Many children in the family
Single (teenage) mom
Stepfamilies
Parental age
Parental SES Parental psychopathology
Community factors
Neighbourhood SES
Neighbourhood violence
Urban vs. rural area
Social cohesion
Societal factors
ISocial factors Cultural factors
Income inequality
War
Natural disasters
Terrorist attacks Social cohesion
Norms or values (honor culture)
Violent media
Results for Objectives from our work package Dissemination
Objective 7.1: Definition of a recognizable and global project identity : GenerACTION.
A Project’s Logo: The project logo was designed and appointed in all the presentations and dissemination material produced and shown by the consortium. The consortium released banners, headers, standard “deliverable template” and a standard “presentation template” for uniform the communications documents and a standard “poster template”.
The project’s official website is available at: http://www.action-euproject.eu. It consists of a public area and a restricted section and contains data protocols, news, events, media, presentations, photo galleries, partners and staff’s information. The website has been constantly updated and extended during the project. The website’s traffic has been constantly monitored with the help of statistic tools. Overview of the visits during the whole project:
-More than 24000 page views and More than 5500 unique users
Sessions to the website are constantly growing from June 2015 to June 2019 (the website was released on December 2014, the monitoring started on June 2015).
-Geographical distribution of users.
Countries: 125 Countries reached. USA is the country with the highest number of users, meaning:
- the ACTION project is currently having a good impact also out of Europe (and also out of the countries with more effort in dissemination);
- research studies on aggression in children are particularly relevant in the USA.
-Cities: 1250 Cities reached. New York and Paris are the (known) cities with the highest number of users, despite no partners of the consortium are based in New York or in France.
Project presence in the social media channels.
The presence of the project on youtube has been ensured with a series of Interviews and videos.
15 videos from the ACTION final meeting (13 Interviews, 1 general video of the event, 1 video containing an award ceremony: best talk given by PhD candidates during the ACTION final event), May 6-8, Sardinia, Italy.
The interviews are available here: http://www.action-euproject.eu/content/videos-and-interviews-action-final-dissemination-meeting
4 Interviews to ACTION staff members and a total of 22 videos showing the ACTION logo. ACTION was a sponsor of the "13th International Workshop on Neonatology", Cagliari, Italy, October 25-28, 2017. A video gallery in the official Youtube page of the event is available and contains the following 4 videos focusing on ACTION, and other videos showing the ACTION Logo.
Interview with Prof. Dorret Boomsma - Main results of the ACTION project: https://youtu.be/6jmwszJXMLM
Interview with Prof. Fanos - The ACTION project's goals
https://youtu.be/PL5NiepwWto
Interview with Dr. Mirko Manchia - Risk factors for aggression
https://youtu.be/ewbbh-UoHU4
Interview with Prof. Luigi Atzori - Metabolomic approach and aggression, in Italian language: https://youtu.be/-9SOaTicUVM
3 Interviews to ACTION staff members and a total of 11 videos showing the ACTION logo. ACTION was a sponsor of the "14th International Workshop on Neonatology", Cagliari, Italy, October 24-27, 2018. A video gallery in the official Youtube page of the event is available and contains the following 3 videos focused on ACTION, and other videos showing the ACTION Logo.
The project’s presence on Twitter and Facebook has been ensured through posts and tweets sent by existing social profiles, using the hashtag #actionproject (according with the notes of the Ethical/Scientific Advisory Board “The Board requested however that Wkpg #7 limit the number of social media used (e.g. Facebook) due to the vulnerability of children and possible privacy violations”, the presence on the social media is limited to announcements and news).
Objective 7.2: Update participations and participating agencies InterACTION.
The consortium developed an online restricted area within the official project’s website: it contains contributions, opinions, shared documents and comments of the staff’s members.
The consortium created ten mailing lists for general@ action-euproject.eu (for all the partners, it can be used for general communications regarding the project (e.g. relevant milestones, events, publications, etc.) and for wp leaders, and for WPs
The consortium developed a Newsletter management system within the website and edited 5 Newsletter Issues (4 regular issues and a Special NTR Newsletter ISSUE from the beginning of the project to the present day). The newsletters were released and directly sent via e-mail to a total of 201 subscribers (including stakeholders, students, researchers):
All Newsletters issues are also available for download in the ACTION website (http://www.action-euproject.eu/content/action-newsletters). They have been also shared by the members of the consortium via private e-mail messages to researchers and stakeholders.
Training events for Junior staff and workshops organized by the consortium (lectures given within workshops and conferences managed by other organizations are not included here but mentioned in the related Work Packages or in the objective 7.4). Examples include:
The second ACTION consortium meeting hosted a Workshop titled "Beyond the Genetics of Aggression; From Clinic to Research and Back". It took place in the Old Meeting Room on the first floor of the Trippenhuis Building, Amsterdam, October 21-23, 2015.
“The ACTION project: dissemination”, session IV of the 5th International Conference on Neonatal and pediatric laboratory medicine, within the 11th International Workshop on Neonatology – From the womb to the adult, Cagliari, Italy, October 26-31, 2015. It had more than 1000 participants and was accompanied by many national and international institutions and scientific societies (Italian Ministry of Health and UNICEF, just to name a few).
Early career scientists workshop on Aggression – Nijmegen, The Netherlands, October 31, November 1, 2016. The event was organized following joint efforts from 4 European projects: ACTION, Aggressotype, Matrics, and FemNAT-CD, and was attended by PhD students and postdoc researchers from all four projects.
“Session: childhood aggression” and “Session: childhood aggression II” at the 47th Annual Meeting of the Behavior Genetics Association, Oslo, Norway, June 28 – July 1, 2017. The ACTION joined forces with EAGLE (EArly Genetics and Lifecourse Epidemiology), organized a symposium made up by two sessions to present the project results
Vu University Amsterdam, October 30-31, 2017. This meeting was associated to a Workshop on Aggression studies dedicated to junior researchers which has been held on October 30th. The event started with talks on aggression-related comorbidity and social topics and the afternoon session was devoted to Biomarkers and (epi)genetics topics.
“Session: Family and Molecular Genetic Studies Provide Insights into ADHD, its Comorbidities (Aggression And Auto-Immune Disorders) and Symptomatology” at The 48th Annual Meeting of the Behavior Genetics Association, Boston, USA, June, 2018
Page with videos and photo galleries: http://www.action-euproject.eu/content/videos-and-interviews-action-final-dissemination-meeting . Photo galleries and Albums are available on Facebook.
Objective 7.3: Inform stakeholders and general public about the project PresentACTION
Some actions performed within previous task (7.2) have been conducted in order to satisfy also the objectives of this task (7.3): e.g. Newsletter exploitation, organization of Workshops and training events. Other actions are also described in the next task (7.4)
Interactive tools. The main outcome of task 7.3 regards the creation of interactive tools in order to show to stakeholders and general public, the comorbidities of child aggression with other childhood psychopathologies. These interactive tools are based on the several results of the ACTION consortium that were published or in press. These tools are also based on several studies and backgrounds of the consortium's partners. These tool have been promoted through publications, abstracts, posters, lectures, newsletters. (see previous technical reports for details).
A first interactive tool, Comorbidity of child aggression tool (based on M. Bartels, et al., “Childhood aggression and the co-occurrence of behavioural and emotional problems: results across ages 3–16 years from multiple raters in six cohorts in the EU-ACTION project”, European Child {&} Adolescent Psychiatry, 2018) shows comorbidities of child aggression, displaying results from questionnaires obtained filled by multiple raters, in six cohorts in the EU-ACTION project”. It can be browsed by questionnaire (ATAC - CBCL - DCB - SDQ - MSNI) or by psychopathology and it is available at this url:
http://www.action-euproject.eu/ComorbidityChildAggression
A second interactive tool, “Teacher Ratings Child Aggression tool”, (based on A. Whipp et al., 2019, “Teacher-rated aggression and co-occurring behaviors among schoolchildren”, paper under preparation) has the overall objective of characterize the levels and associations of aggressive behavior and co-occurring behaviors in the school setting, by using teacher ratings of children at ages 7–14 from population based cohorts of children from Finland, the Netherlands, and the UK.
It is made up by two parts and each of the two parts, similarly to the previous tool, can be browsed by questionnaire (ATAC - CBCL - DCB - SDQ - MSNI) or by psychopathology:
http://www.action-euproject.eu/TeacherRatingsChildAggression
http://www.action-euproject.eu/TeacherRatingsChildAggressionCorrelation
Media coverage targeting the general public. General public is another target of the objective 7.3. During these years the general public has been targeted through the official website, the social media channels (Youtube in particular). The public has been targeted also through articles in online magazines quoting the project and announcing events related to ACTION project. E.g.:
25/10/2018 - Il mondo della Neonatologia a Cagliari, Andrea Congia, UnicaRadio.it
25/10/2018 - Il mondo della Neonatologia a Cagliari, Copenaghenhouse.it
25/10/2018 - A Cagliari workshop internazionale sulla neonatologia, Copenaghenhouse.it
08/05/2017 - Aggression aetiology unveiled, new treatments on the way - Cordis website (available in 6 languages, ACTION is quoted at the end of the article)
27/10/2015 - Press release and and Interview to the Italian tv Videolina
20/10/2015 - 28-minutes audio interview with Prof. Robert Plomin for BBC Radio 4's program - The life Scientific, BBC Radio4
2015 – 40-minutes audio interview with Prof. Robert Plomin - The Guardian newspaper
13/05/2014 - Start of FP7 project ACTION: a study into the complex gene-environment interplay of children with externalizing behavior problems, Emgo.nl
2014 - ACTION - Aggression in children, Healthcompetence.com
2014 - ACTION - Aggression in children, scholarshipdb.net
Objective 7.4: Inform scientists about the project CommunicACTION
Some actions performed within previous tasks (7.2 7.3) have been conducted in order to satisfy also the objectives of this task (7.4): e.g. Newsletter exploitation, organization of Workshops and training events, Release of Interactive tools
The consortium published more than 90 scientific articles (Journals and conference proceedings), Open access has been ensured) focused or related to the ACTION project.
The complete list is available in the ACTION website at the url: http://www.action-euproject.eu/biblio
The consortium participated to a series of symposia at conferences, workshops ,summer schools and scientific meetings. These activities are a key aspect of dissemination and exploitation efforts undertaken within ACTION Work Package 7. The major objectives of these activities are the following:
To ensure a clear understanding of the project and the overall objective, effective knowledge transfer, an integrated approach;
To communicate and share the project’s results and achievements with other researchers;
To receive feedback by professional communities;
To increase the partner network of stakeholders;
To ensure information of and influence on media users and to inform politicians, policy makers, the general public and target groups about the benefits that the ACTION Project can provide;
To utilize the feedback to improve the quality of project results;
To collect new ideas and identify opportunities for further collaborations and funding either as parallel or follow up project activities.
The following list reports about lectures and presentations given by the consortium staff within events managed by other organizations, while the events (workshops and conferences, sessions within workshops) organized or co-organized by the ACTION consortium are listed in the Objective 7.2 of this document and the related presentations (since each event comprises several talks by the ACTION staff members) are better detailed in the Deliverable 7.2.
Objective 7.5: Creation of an updated list of representatives of relevant stakeholders and en-user groups in an interactive stakeholders platform / T7.5 RealizACTION
The consortium has progressively created a liaison database of researchers and stakeholders, with the purposes described in the DoW and in the previous objectives. The consortium organized and participated to a series of symposia at conferences, summer schools and scientific meetings. A selection of contacts comprises :
- Liaison and synergies with other EU funded consortia
Aggressotype consortium
Matrics consortium
FemNAT-CD consortium
CAPICE consortium
- Organizations and institutions
Netherlands Institute for the Study of Crime and Law Enforcement (NSCR)
Union of European Neonatal & Perinatal Societies (UENPS)
European Academy of Paediatric Societies (EAPS)
European Association Perinatal Medicine (EAPM)
Union of the Middle-Eastern Countries and Mediterranean Pediatric Societies (UMEMPS)
Italian Society of Pediatrics
Italian Society of Neonatology
Hellenic Society of Neonatology
SI-DOHAD, Italian Society Developmental Origins of Health and Disease
International Federation Clinical Chemistry – IFCC
Unicef
National Institutes of Health (NIH), U.S. Department of Health & Human Services
Behavior Genetics Association (BGA)
International Society Twin Studies (ISTS)
European Child and Adolescent Psychiatry (ECAP)
Consortium on Individual Development (CID)
Exploitable results: The European ACTION program will achieve its desired impact by ensuring that the project obtains the required visibility through targeted communication measures, and by thoroughly planning dissemination and exploitation of results. Communication efforts will target scientific and other stakeholders, patient groups, counselors, teachers and social workers, and public health service stakeholders. Core activities to maximize the long-term impact of ACTION will ensure that:
The public website www.action-euproject.eu will be kept online for at least 1 year after the project end. This deadline will be almost surely extended to year 2022.
Scientific outcomes will be widely disseminated;
Project results published in peer-reviewed journals also after the end of the project;
Open access to scientific publications will be provided as early as possible and will also be accomplished through the PubMed Central reference number (PMCID) which is a unique number assigned to a work that is posted to PubMed Central (PMC), a free digital archive of biomedical and life sciences journal literature at the U.S. National Institutes of Health;
Project visibility will be created among relevant target groups.
Target audience: In this Section we focus on the possible targets, in terms of public and private institutions, researchers and general public, which may take advantage from news and research products of the ACTION project, after the project end. Several targets include: Scientific community, researchers , higher education, patient organizations, sychiatrists, pediatricians, health care workers, parents, teachers, policy makers
Target’s potential needs related to the achieved results of the project: The ACTION program will advance globally excellent research, which holds the promise of theoretical, methodological, and empirical innovation. It also seeks to promote research based knowledge and insight which has societal, practical and policy relevance. The novel contribution of ACTION will come from the emphasis on: the interactions and interdependencies between different levels of influence on childhood aggression. Rather than treating them as independent, ACTION adopts an interdisciplinary approach to understanding the relationships between different levels of risk and protective factors, focuses on different population groups, and considers the intersections between them to inform various stakeholders and policy development.
Potential Impact:
Potential impact and the main dissemination activities and the exploitation of results.
Some very relevant conclusions can be made about the collection of biomaterials in a clinical setting and its implications for future research.
Firstly, the willingness of parents and children to collect biomaterials did receive more resistance than expected. To study this resistance an inventory among patients refusing participation was conducted. The results of this inventory could function as a guide to inform participants on topics that they deem sensitive as well as maximizing the participation rates in this type of research. Furthermore, during data collection the severity of psychiatric problems of the clinical population increased considerably, which also resulted in lower participation rates.
Although the realization of biological samples was more difficult than expected, there were several major advantages compared to studies focusing on clear-cut specific diagnoses as well as population-based samples. The standardized collection of biomaterials in a broad defined psychiatric sample are guaranteed to have significant problems according to multiple sources (parents school, referrer, and clinicians at the psychiatric facility), enabling generalization to children that actually experience significant impairment from their problem behaviors. In contrast, research in population registers is most of the time reliant on a distribution of problem behaviors, not actual impairment. Additionally, recruitment of children that are referred to psychiatric institutions are screened extensively, therefore very detailed information is available. Importantly, this information is most of the time far more extensive than information which is collected in conventional research (which mostly focuses on parent-reported questionnaires).
Furthermore, the infrastructure for data collection which has been set up thanks to ACTION is still intact, with more participants being recruited to this day. Additionally, the ACTION consortium also realized the contacts with experts on biomarkers and (epi)genetics, which means that future research on biological aspects in a psychiatric setting can be realized more easily than before.
The results with respect to guidelines and clinical practices on severe behavioral problems (SBPs) in children as perceived by research and practice experts, help closing the gap between evidence-based practice and practice-based evidence.
With regards to guidelines, both academics and clinicians share a range of concerns and critical needs around their applicability. Both indicate that familiarity with and implementation of guidelines are modest. Questions around how guidelines are defined and disseminated in practice are also highlighted. Both groups of participants indicate that guidelines should be better adjusted to clinical practices, by expanding on the pool of evidence-based methods included, offering more specific recommendations based on a range of risk factors (e.g. comorbidity, severity, family dynamics, wider systems), taking a multi-factorial approach to formulation and treatment, and increasing focus on systemic working. Merging experts’ and clinicians’ perspectives on guidelines has also indicated insightful gaps in implementation. To this end, our findings inform that more efforts should be directed at persuading commissioners to prioritize SBPs, investing in additional staff training in recommended evidence-based methods, securing national government/ health authorities support, formatting guidelines in a concise and user-friendly format, translating recommendations across countries, and annexing dissemination guidelines. An action plan on key critical needs offers a framework of needs and foreseen challenges in improving guidelines fitness-for-purpose that could be used by guideline development groups when revising current or developing new guidelines. The project also instills a momentum for further development of national or European frameworks for managing SBPs, promoting learning between countries and researchers and clinicians. With regard to treatment practices, our findings provide an initial summary of available prevention and intervention programs across Europe, emphasizing existing gaps in service provision which could further be addressed and explored. The findings also highlight potential barriers to accessing treatment. Finally, by gathering an overview of common and different opinions amongst clinicians concerning specific diagnostic and treatment methods or elements used, the study offers first-hand perspectives on clinical judgment around what intervention is being used for each family and each child and what is perceived as helpful for SBPs. This overview can further provide practice-based evidence to inform guidelines and policy development. From a research perspective, these projects emphasize the value of guidelines in enhancing a shared framework for conceptualizing SBPs to reduce research-practice gaps. The broad overview provided may also inform future more systematic evaluations of highlighted issues. Finally, it promotes sharing of information and expertise between countries.
We found a relation between the age at diagnosis of problem behaviors related to aggression and distal and established that a later diagnosis predicts a higher probability of having a criminal record after age 15 and a lower income as a young adult. Age at time of diagnosis did not affect grade scores. The effect of age at time of diagnosis on suicide attempts was not significant after correcting for multiple testing. Importantly, we showed that the predictive effects of the timing of when the diagnosis occurred on later income and criminal records are independent of effects of background variables that were included in the analyses (i.e. mother’s education and criminal records, neighborhood deprivation, family income, ADHD and ODD diagnoses). The effects were very similar for boys and girls. Our results point in the direction that an early diagnosis could provide the basis for intervention and treatment settings, which may influence the developmental trajectory into later criminality. While very few larger-scale studies have investigated the specific effects of timing of interventions for any neurodevelopmental or disruptive disorder, early intervention has been shown to be effective, especially in children with high initial risk. Our results also support that early intervention has a positive effect on long-term outcomes related to criminality. The effects are in line with the understanding that neural plasticity in the developing brain makes it particularly susceptible to a broad range of interventions. However, given the fact that our data did not contain information on the start of any treatment, some caution is warranted. For instance, it could also be possible that a later diagnosis is merely indicative of persistence, which in turn might be related to poorer outcomes.
Our the results indicate that it is important to monitor aggressive boys and girls not only for future externalizing problems, but also for future internalizing problems. Prevention of specific outcomes, however, may require different interventions. Regarding medication, in young men who received second generation antipsychotics, rates of violent reoffending were lower during periods when individuals received this compared with periods when they did not. Similarly; among released prisoners in Sweden, rates of violent reoffending were lower during periods when individuals were dispensed antipsychotics, psychostimulants, and drugs for addictive disorders, compared with periods in which they were not dispensed these medications. This is valuable information, although further research is needed to understand the causal nature of this association. Finally, SSRI use is associated with elevated rates of violent crime, which remain throughout treatment and persist after discontinuation. Despite the rarity of the outcome, our findings suggest the need for clinical awareness of the risk for violent crime while taking SSRIs, and for provision of information to present and future users of this medication.
The genetic architecture, including the longitudinal genetic architecture, of aggressive behavior in childhood is comparable between countries even with the use of different instruments. An important message for parents, teachers and counselors is that aggressive behavior is heritable and the genes are a main source for differences between children in aggressive behavior. Given the stability of genetic effects it is possible to use phenotypic data collected by the different partners across a large age range. Psychiatric, social, and parenting interventions for Antisocial personality disorder (ASPD) prevention should focus on children and adolescents with high aggression levels, with an aim to gather information from multiple informants.
Aggression co-occurs with almost all other behavioral and social problems. More specifically, aggression co-occurs with oppositional and ADHD-related problems, and at later ages with rule-breaking. In addition to the high correlations of aggression with externalizing problems, we also observed substantial associations with anxiety–depression and other internalizing symptoms. Both for externalizing and internalizing problems, the patterns of co-occurrence were largely gender and rater independent, and were similar even when aggression and the other psychopathologies were assessed by different assessment instruments. Also, there were large similarities in co-occurrence patterns across countries in the Northern part of Europe. The finding that aggression co-occurs with nearly all other behavioral, emotional and social problems during childhood should put aggression in the center of attention, both for research and clinical perspectives.
When causes of differences in aggression during childhood are better understood, this information aids in the development of prevention and intervention strategies. Across definitions, conclusions as to whether different AGG instruments measure the same construct differ. Item content suggests limited overlap, whereas genetic analyses suggest shared etiology among instruments. Whether researchers regard agreement between instruments as satisfactorily depends on their application. It is highly recommended to consider multiple metrics of similarity to decide whether differ measures assess the same. By leveraging a genetically informative design and several commonly used instruments, we attempted to provide a holistic perspective on nuances involved in measurement of AGG in childhood.
The largest genetic association study to date of childhood AGG identified no genome-wide significant associations at SNP level, but identified two genes that are significantly associated with AGG in a gene-based analysis: PCDH7 (protocadherin 7) on chromosome 4p15.1 and ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase 3) located on 1p34.1. There was significant heritability based on single nucleotide polymorphisms. Genetic correlations were high and positive for a series of disorders and behaviors including ADHD, Bipolar disorder, Loneliness, Major depression, Risk tolerance, and Smoking. Genetic correlations were high and negative for educational attainment and age of first child. Stratified analyses revealed significant differential genetic correlations between AGG and other behavioral, emotional and social problems, and life outcomes; highlighting the importance of combining information from multiple sources to get the most complete overview of the genetic etiology of childhood AGG.
Within Action, we have generated a large amount of biomarker and ‘omics data to study childhood aggression. We developed and piloted extensive protocols for sample collection, i.e. urine and buccal-cell collection in children, which were made available for the scientific community (http://www.action-euproject.eu/content/data-protocols) for future projects.
In ACTION we report the first findings of the genetic association studies, the epigenetic projects and the metabolomics studies. All genome-wide results and all products from the metabolomics projects are made available to the scientific community, to not only aid in the understanding of childhood aggression, but also to study the links with adult outcomes through techniques that do not require the raw data, but can take summary statistics to estimate genetic correlations, construct polygenetic scores for risk prediction, based on aggression and attention problems discovery studies and on the extensive database for SNP-metabolomics associations, because we have also undertaken a large-scale investigation of the genetic architecture of blood metabolites. As part of this undertaking we created a database of all published (2008-2018) metabolite-SNP associations, which includes characterization and classification of all metabolites. This database has been made publicly available (http://bbmri.researchlumc.nl/atlas/#data). Overall, ACTION provides tools for future scientific research, ranging from potential amino acid intervention studies (protocol), to polygenetic prediction or Mendelian Randomization studies (SNP-metabolite associations) and multi-omics investigations (ACTION ‘omics data).
For Action, a steroid hormone platform to be used for urine samples, was developed which offers great opportunities to the scientific community to explore the potential relationship of hormones and (subtypes of) aggression and other behaviors and behavioral problems in children.
The extensive findings in Action reflect collaborative efforts to integrate genetic, genomic and multilevel environmental methods to better characterise aggression in childhood. Details of the advanced statistical modelling and study results and implications have been communicated internationally in order to widen the impact and public understanding of the genetic and environmental antecedents to aggression. Through high impact publications, media coverage and invited lectures and conferences, we have collectively raised the research profile of childhood aggression and demonstrated the utility of large genetically sensitive collaborative efforts in studying its etiology. Crucially, our multi-method approach to investigate GxE and rGE using twin, polygenic score, epigenetic and atheoretical machine learning has produced exciting preliminary results which are currently being replicated in the ACTION cohorts and will promote future dissemination activities even after the end of the ACTION grant period.
In collaborative projects, it is not uncommon that research groups use different instruments to assess a trait. We examined convergence between different instruments to assess aggressive behaviors, focusing on item content, clinical concordance, correlation, and genetic overlap. The findings revealed that agreement between different measures of childhood aggressive behaviors depended on the metric of agreement under consideration (i.e. item content, clinical concordance, correlation, underlying genetics). Overlap between the item content of the aggressive behavior measures was absent to moderate. Concordance on who received a score above the clinical cut-off was very weak to weak. Associations between the different measures of aggressive behavior based on continuous scores, while correcting for skewness, yielded higher agreement (i.e. moderate to strong) than clinical cut-off scores. Genetic correlations ranged from weak to very strong, which generally indicated high overlap in underlying genetics between the different measures of aggressive behavior. From this study, two patterns emerged that may impact future practice and research. First, higher convergence between instruments for correlation than for clinical concordance suggests that decision-making based on continuous scores instead of clinical cut-offs may improve reliability. Second, the generally high genetic overlap suggests that the same genetic factors influence the different measures of aggressive behavior, despite lower convergence in item content, clinical concordance, and correlation. Therefore, the findings indicate that different measures of aggressive behavior can readily be combined in future collaboration studies on the genetics of childhood aggressive behavior.
We also provided a harmonized aggression score that facilitates future collaborations to further investigate the etiology of childhood aggression. This work allowed for a more generalized statement about aggression in school-aged children from the Netherlands, Finland, Sweden, and the United Kingdom. The harmonized aggression score permitted analyses to be conducted in a large international sample in which all participants had a comparable score. The harmonized aggression score is available to ACTION partners for future joint analyses. Moreover, the reference panel created for this study may facilitate future linking and scaling of phenotype scores across and beyond the consortium. The integrated data analysis framework discussed in the study provides promise for future multi-cohort collaborations to investigate behavioral or psychological phenotypes. Often different cohorts assess traits using different items which introduces measurement heterogeneity. The method presented in this study should remove this specific type of heterogeneity and consequently benefit future joint analyses of multiple studies.
Both genetic and environmental factors are important contributors to individual differences in childhood aggression and a comprehensive understanding of the etiology of childhood aggression requires taking both genetic and environmental factors into account, as well as examining the possible interplay between them. A socio-ecological framework was considered that acknowledges the importance of biological and psychological systems at the individual level together with factors at distal levels including parents and the family, the community, and finally the geopolitical level.
At the distal level, there is a negative association between national-level policies for child and adolescent mental health (CAMH) and adolescent aggressive behavior (N = 172,829, from 30 European countries). This indicated lower aggressive behavior in countries with more CAMH policies. National-level policies in this study consisted of measures of availability of epidemiological data, the number of CAMH facilities, investment in family benefits, and investment in education. We obtained information on these policies from renowned statistical institutes (e.g. Eurostat, 2016; Institute for Health Metrics and Evaluation, 2018; OECD Social Policy Division, 2016). This study was the first to examine the association between CAMH policies and adolescent aggressive behavior, and revealed the potential of CAMH policies to function as large-scale prevention to ameliorate adolescent aggressive behavior. Furthermore, this study provided directions for future research to better understand the mechanisms underlying this association.
On a more proximal level, we aimed to predict childhood aggression with a wide range of family- and individual-level predictor variables. The outcome measure was physical overt aggression, assessed around age 9 and harmonized across multiple European cohorts. Application of rigorous methods that permitted simultaneous analyses of all included predictor variables yielded a multi-group model that accounted for interactions with sex and cohort and confirmed linear main effects of variables measuring non-physical aggression, ADHD and conduct disorder related behaviors, maternal smoking during pregnancy, parenting, and proportion of life events. The results provided a set of easily observable behaviors that might help to identify children more likely to become aggressive. Simultaneous modelling of variables yielded insight in their importance when also taking other variables into account. An additional benefit from the rigorous methods was that they allowed for testing different types of predictor effects (i.e. linear, non-linear, interaction) without having to pre-specify them. These methods may provide tools for future (collaborative) research projects that aim to predict a trait using a wide range of variables .
The contribution of genetic and environmental factors to individual differences for a trait may vary as a result of different backgrounds (i.e. gene-environment interaction), such as levels of socioeconomic status (SES). We examined the moderating effect of SES on the genetic architecture of childhood aggression in large samples of 7-year-old twins SES moderated the contribution of genetic and environmental factors. The contribution of genetic factors was similar across SES strata, the contribution of shared environmental factors was lower, and the contribution of nonshared environmental factors was higher for children from a high SES background compared to children from a low or medium SES background. This pattern was similar for children from the Netherlands and the United Kingdom.
This finding indicates that children from low or medium SES backgrounds would benefit more from treatment to improve shared environmental factors. Examples of such shared environmental factors are healthy family functioning, less parental stress, housing quality, or positive school attachment. To some extent, policies for child and adolescent mental health such as investment in family benefits and investment in education attempt to already do so through alleviation of factors associated with low SES. The results suggest that such early prevention measures on the shared environment indeed may benefit children at risk for heightened levels of aggression.
The literature synthesis of 72 systematic reviews and meta-analyses on treatment effectiveness for childhood aggression revealed that effect sizes were mostly absent or small for universal and selective prevention, and mostly small to medium for indicated prevention and intervention. For moderators including child age, child gender, implementation to individuals or groups, person implementing the treatment, different treatment programs, and session related factors or treatment intensity, these effect sizes were invariant in a majority of studies. Moderator effects were mixed for socioeconomic status, type of treatment, informant, and research quality. The only two significant moderators comprised of pre-treatment levels of aggression and parental involvement. Treatment effectiveness was higher for children with higher levels of aggression before treatment and when parents were involved in the treatment.
A first pattern that emerged from the literature synthesis was that children with sub-clinical levels of aggression appeared to benefit as much from treatment as children with clinical levels of childhood aggression. Secondly, results revealed that based on existing research it is not yet possible to distinguish subgroups of children that would benefit more from treatment for aggression than others. The positive moderating effect of parental involvement on treatment effectiveness for childhood aggression suggests that an opportunity for future research may be to focus more on parental influences as possible moderators of treatment effectiveness. These findings provide additional support that it would be beneficial to screen for parental psychiatric disorders for the treatment of childhood behavior problems .
The wealth of available data and expertise within ACTION permitted examination of new research questions. The studies provided tools for the assessment of childhood aggression (and other traits) in future collaborative research projects. Assessing whether children score above or below a clinical cut-off for inclusion for treatment may lead to children to be excluded from treatment from which they would benefit as much as children who score above a clinical cut-off. Moreover, measures of aggressive behavior agree only to a small extent on which children display clinical levels of aggression, which may cause children to miss out on treatment not because of their level of aggression, but because of the measure selected to assess their aggression. A harmonized factor score of aggression may facilitate future collaborative research projects. In addition, the research in WP6 highlighted the complexities in the etiology of childhood aggression. Childhood aggression is found to be associated with a broad range of factors, from country-level policies to more proximal factors as the family environment, and individual level factors such as behavior and genetics, which may also interact.
Although the influence of the broad range of factors we considered on childhood aggression adds complexity to the etiology of childhood aggression, it also provides opportunities to improve prevention and intervention strategies for childhood aggression. For example, inclusion of parental characteristics (i.e. a family based approach) in diagnosis and treatment might improve treatment effectiveness for childhood aggression. Additionally, more policies for child and adolescent mental health were associated with lower levels of aggressive behaviors, which suggests merit in employment of policies as early prevention efforts. Moreover, differences in etiology of aggression as a result of socioeconomic background highlight that it is promising to distinguish subgroups of children more likely to develop childhood aggression and children more likely to benefit from treatment. The research contributes to previous work to advance our understanding of assessment, etiology, and treatment of childhood aggression and provided directions for future research working towards a more personalized approach to childhood aggression.
Action has been very involved in all dissemination activities with a very accessible website, movies newsletters and interviews on Youtube channels, and a great set of interactive tools.
Digital presence:
Website: www.action-euproject.eu
Youtube channel: https://www.youtube.com/watch?v=j0Ev8h1ARkE
Hahshtag on Facebook and Twitter: #actionproject
Comorbidity of child aggression tools:
http://www.action-euproject.eu/ComorbidityChildAggression
Total number of interactive charts: 25
- Teacher Ratings Child Aggression tool
http://www.action-euproject.eu/TeacherRatingsChildAggression
http://www.action-euproject.eu/TeacherRatingsChildAggressionCorrelation
Total number of interactive charts: 35
Videos, interviews :
o ACTION official video: https://www.youtube.com/watch?v=j0Ev8h1ARkE
Multiple Videos from the ACTION final meeting on aggression studies, May 6-8, 2019 Pula, Cagliari, Sardinia, Italy
- Childhood aggression and its comorbidities: overview of the dissemination meeting: https://youtu.be/fFPXU9Hmrpw
- Prof. Meike Bartels: main results of the ACTION project - Childhood aggression and its comorbidities: https://youtu.be/PJBocGFurBo
- Prof. Robert Vermeiren: clinical implications of the ACTION project - Childhood aggression: https://youtu.be/_jdtr5I-mio
- Prof. Gitta Lubke: Predicting childhood aggression; mining large datasets followed by confirmatory modeling: https://youtu.be/qKcyqWIh-7Q
- Dr. Jenny van Dongen: Epigenome-wide association study meta-analysis of aggressive behavior: https://youtu.be/y7HUrl9oZow
- Hill Ip, PhD candidate: Genome-Wide Meta-Analysis of Aggressive Behavior across Rater, Instrument and Age: https://youtu.be/wpDax_Hiq9I
- Dr. Margherita Malanchini: aggressive behavior in childhood and the role of smoking during pregnancy: https://youtu.be/w2oQ6E0uncc
- Yayouk Willems, PhD Candidate: Environmental and Genetic contributions to low self-control: https://youtu.be/lg21SEugGeM
- Peter Roetman, PhD Candidate: aggression in clinical practice: https://youtu.be/Xz8Qs9ZYxzI
- Andrea Allegrini, PhD Candidate: Multi-trait polygenic prediction of broad aggression in adolescence: https://youtu.be/DUApzodNyEg
- Dr. Lucia Colodro Conde: a twin study of oppositional defiant disorder behaviors using data from Australia and The Netherlands: https://youtu.be/CMdw6uxfBRs
- Fiona Hagenbeek, PhD Candidate: Urinary biomarkers and metabolites of childhood aggression: https://youtu.be/74zK-zD9-6I
- Anne Hendriks, PhD Candidate: What are the implications of five years of research on childhood aggression? https://youtu.be/22ZSXXmTDNs
- Alyce Whipp, PhD Candidate: Teacher-rated aggression and co-occurring behaviors: https://youtu.be/bf59AxosxII
- Roberta Pintus: Lipids and aggression: what is going on?: https://youtu.be/3Uvaa2YQNwQ
- Childhood aggression and its comorbidities: concluding remarks and award ceremony (best talk given by PhD candidates: Fiona Hagenbeek - Urinary biomarkers and metabolites of childhood aggression): https://youtu.be/PekNHUZLDKU
o Videos from the "14th International Workshop on Neonatology", Cagliari, Italy, October 24-27, 2018.
- Interview with Prof. Vassilios Fanos - session for early researchers within the 14th Workshop on Neonatology:https://youtu.be/jHD1rqnryaA
- Interview with Dr. Mirko Manchia - Clinical implications of the ACTION project
https://youtu.be/RJCOCYLH_8Q
- Interview with Dr. Matteo Mauri - Role of the University of Cagliari in ACTION, dissemination and main results of the ACTION project: https://youtu.be/j1vrt2eW6QM
o Videos from the "13th International Workshop on Neonatology", Cagliari, Italy, October 25-28, 2017.
- Interview with Prof. Dorret Boomsma - Main results of the ACTION project: https://youtu.be/6jmwszJXMLM
- Interview with Prof. Fanos - The ACTION project's goals: https://youtu.be/PL5NiepwWto
- Interview with Dr. Mirko Manchia - Risk factors for aggression: https://youtu.be/ewbbh-UoHU4
- Interview with Prof. Luigi Atzori - Metabolomic approach and aggression, in Italian language: https://youtu.be/-9SOaTicUVM
- Other interviews and videos (11 in total) sponsored by ACTION: official Youtube page of the event
o Other Videos/Interviews (last access 2016)
- Project Kick-off meeting: https://youtu.be/c5b_4dtg8nA
- Interview with Robert Plomin in Times Education Supplement - Genetics and Education: https://youtu.be/-4VCISnwE-k
- "A new Scientific Humanism: Seven Chapters on Narrative Medicine and Precision Medicine", Florence 18-19 February 2016: http://www.action-euproject.eu/content/new-scientific-humanism-seven-chapters-narrative-medicine-and-precision-medicine-florence-18
- Webinar at NIH - National Institutes of Health, U.S. Department of Health & Human Services: Application Metabolomics to Provide Pediatric Biomarkers: https://youtu.be/URVBX57qUdI
- Interview with prof. Vassilios Fanos (Italian TV Videolina): http://www.action-euproject.eu/content/interview-prof-vassilios-fanos-italian-tv
Newsletters (all available on website)
ACTION Newsletter N°1, December 2015;
ACTION Newsletter N°2, August 2016;
ACTION Newsletter N°3, July 2017 ;
ACTION Newsletter NTR SPECIAL ISSUE, September 2017;
ACTION Newsletter N°4, March 2018;
ACTION Newsletter N°5, ACTION Final meeting invitation, February 2019 (not available for download)
List of Websites:
public website: http://www.action-euproject.eu/
contact details: n.stroo@vu.nl