European Network of National Schizophrenia Networks Studying Gene-Environment Interactions
6200 Md Maastricht
Higher or Secondary Education Establishments
€ 1 340 958,26
Laurent Louwies (Mr.)
Sort by EU Contribution
KING'S COLLEGE LONDON
€ 1 031 012,74
€ 2 798 872
ZENTRALINSTITUT FUER SEELISCHE GESUNDHEIT
€ 776 617
€ 1 042 418
SERVICIO MADRILENO DE SALUD
€ 1 181 892
Academisch Medisch Centrum bij de Universiteit van Amsterdam
€ 1 071 201
INSTITUT NATIONAL DE LA SANTE ET DE LA RECHERCHE MEDICALE
€ 350 560
KATHOLIEKE UNIVERSITEIT LEUVEN
€ 414 000
UNIVERSITY MENTAL HEALTH RESEARCH INSTITUTE
MEDIZINISCHE UNIVERSITAET WIEN
€ 108 915
€ 109 171
KLINIKUM DER UNIVERSITAET ZU KOELN
€ 112 905
€ 75 500
€ 175 405
E.C.S. INTERNATIONAL BV
€ 163 154
Omega Pro Proje Arastirma Gelistirme ve Danismanlik Ltd Şti
€ 40 000
€ 53 172
THE UNIVERSITY OF HONG KONG
€ 179 940
UNIVERSIDAD DEL PAIS VASCO/ EUSKAL HERRIKO UNIBERTSITATEA
€ 184 998
THE CHANCELLOR MASTERS AND SCHOLARS OF THE UNIVERSITY OF CAMBRIDGE
€ 194 940
ROYAL COLLEGE OF SURGEONS IN IRELAND
€ 54 000
€ 12 000
ALMA MATER STUDIORUM - UNIVERSITA DI BOLOGNA
€ 73 224
UNIVERSITA DEGLI STUDI DI PALERMO
€ 72 000
UNIVERSITY OF MELBOURNE
Grant agreement ID: 241909
1 May 2010
30 April 2015
€ 15 030 922,89
€ 11 616 855
Schizophrenia risk factors
Grant agreement ID: 241909
1 May 2010
30 April 2015
€ 15 030 922,89
€ 11 616 855
Final Report Summary - EU-GEI (European Network of National Schizophrenia Networks Studying Gene-Environment Interactions)
The aim of EU-GEI is to identify the interactive genetic, clinical en environmental determinants involved in the development, severity and outcome of schizophrenia. In order to identity these interactive determinants, EU-GEI has employed family-based, multidisciplinary research paradigms, which have allowed for the efficient assessment of gene-environments interactions. In order to go beyond old findings from historical convenience cohorts with crude measures of environmental factors and clinical outcomes, the focus in EU-GEI has been on recruitment of new family-based clinical samples with state-of-the-art assessments of environmental, clinical, genetic determinants as well as their underlying neural and behavioural mechanisms. New statistical tools have been developed to combine the latest multilevel epidemiological with the latest genome-wide genetic approaches to analysis. Translation of results to clinical practice has been facilitated by additional experimental research and risk assessment bioinformatics approaches. This has resulted in the identification of modifiable biological and cognitive mechanisms underlying gene-environment interactions and the construction of Risk Assessment Charts and Momentary Assessment Technology tools with can be used for (i) early prediction of transition to psychotic disorder in help-seeking individuals with an at-risk mental state and (ii) early prediction of course and outcome after illness onset. In order to reach these goals, EU-GEI has assembled a multidisciplinary team of top schizophrenia researchers who have the range of skills required to deliver a program of research that meets all the call’s requirements and who have access to collect a number of unique European samples. The partners in EU-GEI represent the nationally funded schizophrenia / mental health networks of the UK, Netherlands, France, Spain, Turkey and Germany as well as other partners.
Project Context and Objectives:
The aim of EU-GEI is to identify, over a 5-year period, the interactive genetic, clinical and environmental determinants, involved in the development, severity and outcome of schizophrenia. In order to identify these determinants and their interactions, we will employ family-based, multidisciplinary study paradigms, which allow for efficient assessment of gene-environment interactions. Translation of results to clinical practice is facilitated by additional experimental research and risk assessment bioinformatics research. This enables (i) the identification of modifiable biological and cognitive pathways and mechanisms and (ii) the construction of Risk Assessment Charts and Momentary Assessment Technology tools which can be used for the early prediction of (i) transition to psychotic disorder in at-risk help-seeking individuals, (ii) diagnosis and (iii) outcome monitoring.
In order to reach these goals, EU-GEI has assembled a diverse multidisciplinary team of top schizophrenia researchers who have the required range of skills and facilities. The partners in EU-GEI represent the nationally funded schizophrenia / mental health networks of the UK, Netherlands, France, Spain, Turkey and Germany as well as other partners. EU-GEI additionally includes a user-led PhD project, and has aimed to impact on mental health and societal issues as called for in the European Parliament Resolution on Mental Health in Europe.
The Concept of GxE in Schizophrenia
GxE Research in Schizophrenia: Time for Serious Consideration?
The high heritability of schizophrenia provides a critically important route of inquiry into the pathophysiology of this devastating disorder. However, attempts to discover common alleles that relate directly to schizophrenia have only very recently been productive and point to very small odds ratios (O'Donovan et al., 2008a) at least when main effects alone are sought. The application of novel genetic technology to schizophrenia has also established that rare fairly high penetrance copy number variants (CNVs) may account for a small proportion of the genetic liability of schizophrenia (International Schizophrenia Consortium, 2008). However, unpublished analyses of the International Schizophrenia Consortium (ISC) SNP data along with other datasets suggest that at least half of the genetic variance results from alleles of small effect (Purcell, presented on behalf of the ISC, ISPG, 2008, Osaka). A separate and concurrent line of research, mainly in the area of epidemiology, has established that there are high rates of schizophrenia in large cities, immigrant populations, traumatised individuals and cannabis users, at least some of which is thought to be the result of underlying environmental exposures. However, the biological mechanisms underlying these established environmental risk factors are largely unknown. Given the fact that factors like urbanicity, drug use and migration are of major importance to the EU, these epidemiological findings needed further clarification.
Exciting findings in other areas of medicine have motivated researchers to turn their attention to better understanding the complex ways in which nature interacts with nurture to produce schizophrenia. This genotype x environmental interaction (hereafter: GxE) approach differs from the linear gene-phenotype approach by positing a causal role not only for either genes or environment in isolation, but also for their synergistic co-participation in the cause of schizophrenia where the effect of one is conditional on the other (EU-GEI, 2008; Sham, 1998). For example, genes may moderate the psychotogenic effects of certain drugs of abuse, or the environment may moderate the level of expression of a gene associated with schizophrenia.
The European Network of Schizophrenia Networks for the Study of Gene-Environment Interactions has argued that systematic attempts to identify gene-environment interactions cannot simply be equated with traditional molecular genetic studies with a number of putative environmental variables thrown in (EU-GEI, 2008). Therefore, EU-GEI has gone go beyond paying lip service to the “stress-vulnerability” model of schizophrenia and has brought many disciplines together to work on the identification of gene-environment interactions.
Several Relevant Gene-Environment Relationships Exist
Genes and environments can impact on human health in several different ways. First, an environmental factor, in interaction with other environmental factors, may represent a sufficient cause of ill-health (E-only) or a gene, in interaction with other genes (epistasis), may represent a sufficient cause (G-only), i.e. each may have separate additive effects. Alternatively, genes may influence environmental sensitivity, or environments may cause genetic (epi)mutations impacting on gene structure and gene expression, giving rise to gene-environment interactions. Thus, in the case of GxE, genes and environment may have a more-than-additive joint effect. Finally, genes may influence exposure to a causal environmental exposure and thus contribute indirectly to the outcome (environment is on the causal pathway from gene to outcome; rGE+). Genes may also contribute to both the environmental exposure and the outcome; in this case, the environment is non-causal (genetic confounding environmental association; rGE-).
GxE Research in Schizophrenia: Need for Translational Approach
Genes cannot be directly manipulated for the purpose of psychiatric treatment or prevention. However, at the level of proximal interactions with environmental component causes that impact on the underlying pathophysiological mechanisms, intervention is possible. Therefore, in order for GxE research in schizophrenia to be translated to the level of prevention and treatment, a focus on environmental mechanisms in GxE is vital. Major challenges remain, however, as currently there has been very little research regarding the mechanisms of proximal interaction with environmental effects. Indeed, much uncertainty remains regarding the role of environmental factors in schizophrenia (EU-GEI, 2008). Therefore, insight into the role of environmental factors in GxE and their neural and cognitive mechanisms was the realistic first step of the EU-GEI project towards translation of results to the clinic. This knowledge may contribute to manipulations of the environment that interact with genetic variation, or to manipulations of the neural and cognitive consequences of environmental exposures interacting with genetic risk. Mechanistic knowledge also facilitates translation of GxE results to the area of early prediction.
GxE Research in Schizophrenia Raises Ethical Issues
In the case of common risk alleles each conferring small effects upon schizophrenia risk, knowledge about individual genes does not provoke ethical dilemmas, as predictive values are very low. However, if the interaction between genes and environment together produce substantial risks, ethical issues may arise if this knowledge is examined for its predictive value. Similarly, recent work suggests that some genetic variants exist that, although rare, are associated with substantially higher risks than the small effect sizes of common variants. Therefore, GxE research in EU-GEI has taken into account the ethical implications and has explored societal issues arising from this. An important issue that has been raised in this regard (Van Os and Delespaul, 2005) is in what type of population predictive tools are to be used. Use of prediction in populations that are not seeking help in any way raises major ethical issues, given the right “not to know” and the stigma associated with mental illness (McGlashan, 2005, Phillips et al., 2005). These issues have been explored further during the project, in particular if, genetic information is to be used to co-predict morbidity trajectories in this vulnerable population (see work package 9: Ethics).
GxE Research in Schizophrenia Should Involve Stakeholder Groups
EU-GEI has consulted with different stake holder groups before and during the project. EU-GEI has consulted with (i) ENUSP, the European Network for Users and Survivors in Psychiatry and (ii) EUFAMI, the European Federation of Families of Persons with Mental Illnes, EU-GEI has included a specific user-led PhD project on stigma as a clinical predictor of course in schizophrenia (see work package 7: Course). EU-GEI has furthermore consulted and obtained support from the European College of Neuropsychopharmacology (ECNP).
GxE Research in Schizophrenia Raises Statistical Issues
Epidemiological research focusing on environmental factors has evolved towards alternative forms of multilevel analyses disentangling molecular, neural, individual and societal effects (Susser and Susser, 1996). Molecular genetic research has evolved towards genome-wide. The statistical challenges associated with these advances are numerous and complex, and have given rise to a large number of statistical innovations (Ao et al., 2005, Curtis et al., 2006, Knight et al., 2008, Neale and Sham, 2004, Purcell et al., 2007a, Purcell et al., 2007b). In genetic epidemiology, gene-environment interaction research is making use of a variety of special designs including case-only, case-control, case-sib, twin, case-parent and family case-control designs. These designs are particularly useful if a priori hypothesized simple molecular genetic variation at the level of a single SNP is modelled. However, it has been pointed out that to recent date, no comprehensive statistical solutions were available for a priori or explorative gene-environment interaction hypotheses making use of complex genetic data at the level of multiple haplotypes or at the level of genome wide marker information (EU-GEI, 2008). Therefore, EU-GEI has developed new statistical tools in order to support its large-scale effort towards identifying gene-environment interactions.
Given the above conceptual issues, EU-GEI has the following S&T objectives:
IDENTIFYING DETERMINANTS AND MECHANISMS:
1. To identify genetic, environmental and clinical determinants of schizophrenia vulnerability, onset, severity and course beyond existing knowledge.
2. To assess neural system and behavioural substrates mediating gene-environment interactions.
TESTING FOR GXE IN CLINICAL SAMPLES AND DEVELOPING TRANSLATIONAL TOOLS:
1. To test for gene-environment interactions, as well as gene-gene and environment-environment interactions playing a role in schizophrenia onset, vulnerability, severity and course.
2. To develop new, family-based statistical tools to combine genome-wide genetic with epidemiological approaches in the analysis of GxE
3. To develop translational tools for the early prediction, diagnosis and course of the disease, focusing on (i) Risk Assessment Charts and (ii) Momentary Assessment Technology
MANAGEMENT, ETHICS, IMPACT AND TRAINING:
1. To create a multidisciplinary training environment for the research of determinants and interactions in schizophrenia, combining knowledge from clinical psychiatry, clinical psychology, genetics, epidemiology, biostatistics, neuroscience and neuroimaging
2. To create value for money by making available data collected in EU-GEI in order to facilitate unencumbered global data sharing between researchers
3. To develop, in collaboration with stakeholder groups, an ethical context for translational GxE research in schizophrenia.
4. To disseminate results to a critical mass of stakeholders for policy counselling, integrating the expertise and views of the European Commission and all relevant stakeholder groups to develop policy recommendations and guidelines
WORK PACKAGE 1: MANAGEMENT
Within the Consortium P1 MUMC has been designated to act as the coordinating centre, with Professor Jim van Os as coordinator and Dr. Bart Rutten as vice-coordinator.
The Executive Board has consisted of the coordinator and all the Work Package leaders (WPs 1-11).
The coordinator and the Executive Board were assisted by the Management Support Team, located in Maastricht at MUMC, consisting of the project coordinator and vice-coordinator, a project manager and a project assistant. Furthermore, the consortium’s database was coordinated and managed by a senior database manager appointed at MUMC.
The Management Support Team at MUMC has ensured the daily management and implementation of EU-GEI activities, financial and administrative matters. Contingency plans have been developed and implemented in order to maximize recruitment and data collection, and progress of the project has been closely monitored.
The first year of the study has been dedicated to selection, development and adaptation of the instruments used for the diagnostic assessment, assessment of clinical characteristics and environmental risk factors. WP1 has organized and coordinated the translation of the instruments in 8 different languages. Translators, knowledgeable of the language of the original version of the instrument, but with original mother tongue of the target culture, were used for the forward-translation. All translators were familiar with schizophrenia research and all terminology of this area. Using the same approach as outlined in forward-translation, an independent translator who had not yet been exposed to the questionnaire before, then translated the instrument back to the original language. This effort has resulted in a unique set of instruments and manuals available in Dutch, English, Turkish, Portuguese, Serbian, German, Italian, French and Spanish (M1-12), which have formed the new European-wide standard and have allowed for ‘uniform’ testing throughout the various EU countries.
The coordination and organisation of the central database has been organized and coordinated by the data manager of WP1. This has included: the preparation of the database for the entry of all EU-GEI data, preparations of guides and manuals for data entry and export, preparation of user accounts for all partners and the organization of monthly skype meetings with different Work Packages 2, 5 and 6. Once the data entry started, WP1 has monitored data entry and supervised data consistency and quality. In the final year of the project, data cleaning, data export and data analyses have been at the forefront of efforts.
Overview Consortium meetings:
1. M1 – Year 1: Kortenberg / Belgium
Kick Off meeting, Executive Board meeting, training session “Train the trainers”, Work Package meeting and General Assembly meeting
2. M19 – Year 2: London / United Kingdom
Consortium meeting, Executive Board meeting, training session “Brush-up training on the use of instruments”, Work Package meeting and General Assembly meeting
3. M31 - Year 3: Palermo / Italy
Consortium meeting, Executive Board meeting, training session “third brush-up training and training-lectures on how to conduct statistical analyses in Gene x Environment interaction studies”, Work Package meeting and General Assembly meeting
4. M43 – Year 4: Barcelona / Spain
Consortium meeting, Executive Board meeting, training day on GxE statistics and publications, Work Package meeting and General Assembly meeting
5. M55 – Year 5: Belgrad / Serbia
Consortium meeting, Executive Board meeting, Work Package meetings, meeting of the General Assembly
To ensure and optimise recruitment and sample size, 7 affiliated partners have been included and have participated in Work Packages 2, 5 and 6. As conditions of participation, it was ensured that in each of the centers the identification, recruitment and assessment of subjects would meet the criteria and standards set for all centers. Furthermore the costs made by the affiliated centers would not be claimed from EU-GEI.
Add-on studies to EU-GEI
The infrastructure of the EU-GEI consortium has resulted in several research proposals that were added on the core EU-GEI research aims, and that will be performed according to consortium rules of EU-GEI. A protocol for submission of add-on study proposals for approval by the Executive Board was established. It has been ensured that the add-on studies would not require any financial resources from the EU-GEI budget. In fact by having the opportunity to use EU-GEI’s infrastructure, key researchers coordinating these add-on studies have had significant successes in obtaining financial resources from other national research foundations for these add-on studies, e.g. by the UK Medical Research Council (to partner IoP), the French national agency for research (to partner INSERM), the Irish Research Council (to partner RCSI) and the Netherlands Organisation for Scientific Research (to partner MUMC).
Continuation after M60
During year 5 a Declaration of Intent has been established, extending the Consortium Agreement (CA). The extension of the CA will serve as an unofficial agreement between EU-GEI partners, governing a number of issues that will continue after the end date of the project. The issues include:
- the completion of the planned tasks set out in the DoW but not yet completed at the official end date of the project;
- the further exploitation of the generated data base as well as use of EU-GEI samples;
- complying with the agreed upon publication rules and procedures;
- acknowledging the structure of the EU-GEI board, consisting of the WP leaders of all WPs.
The PI’s of those EU-GEI partners that will be involved in one or more of the above mentioned activities have signed the Declaration of Intent and commit to the agreed upon rules of cooperation, ensuring the persistence of the successful scientific consortium.
As coordinator of the consortium WP1/P1.MUMC will continue to coordinate ongoing activities after the official end of the project (M60) as long as necessary. The coordination will include ongoing analyses, publication procedure and training facilities: internal and external.
WORK PACKAGE 2: FUNCTIONAL ENVIROMICS
Incidence, environmental risk factors GxG and ExE
EU-GEI WP2 has sought to estimate incidence rates of schizophrenia and other psychotic disorders in 15 sites in 6 countries in order to test hypotheses about variations among and within countries, including by urbanicity (population density) and by migrant and minority ethnic group.
Overall, we identified 2,553 cases from a total of 12,880,210 person years at risk, over an average study period of 36 months, making this the largest study of the incidence of schizophrenia and other psychotic disorders ever conducted.
On the basis of preliminary and unstandardised analyses of crude incidence rates, four findings stand out:
1) Crude incidence rates varied markedly among sites, ranging from 10.0 per 100,000 person years in Cuenca, Spain to 63.9 per 100,000 in London, UK, a six-fold difference.
2) In the northern European countries (UK, Netherlands, France), the incidence was higher urban compared with the rural sites. That is, the incidence in urban sites (London, Amsterdam, Creteil) was 2 to 3 times higher than in the rural sites (Cambridge, Leiden, Clermont).
3) In the southern European countries (Italy, Spain) and Brazil, there was no evidence that the incidence varied among sites.
4) Incidence rates in southern Europe (irrespective of degree of urbanicity) and Brazil (range: 10.0 to 24.4 per 100,000) are similar to or lower than those in rural sites in northern Europe (range: 18.0 to 22.9 per 100,000).
In preliminary analyses, we found marked variations in crude incidence rates among sites. Most intriguingly, rates were generally lower in southern Europe and Brazil and the high incidence in urban sites appears to be limited to northern European cities. There is, however, a need for caution and these initial findings require confirmation in further analyses. In particular, in further analyses we will standardise rates for age and gender, we will more precisely define degree of urbanicity in terms of population density, and we will adjust for migration and ethnicity. If confirmed, these findings raise important questions about the nature of the relationship between urbanicity and psychosis and about how different economic, social, and cultural contexts impact on risk that we will explore in more detailed analyses of WP2 data. In addition, we will extend our analyses to test hypotheses concerning, for example, variations in incidence among migrant and minority ethnic groups.
Environmental risk factors
EU-GEI WP2 was specifically designed to generate, in each site, samples of first episode cases and of population based controls with extensive information on exposure to a full range of environmental factors, including childhood indicators and experiences within and outside the home. This was to allow us to test, in detail and robustly, hypotheses about the impact and interaction of environmental factors across the life course on odds (risk) of psychotic disorder. Here we report preliminary analyses of data on childhood adversities and psychosis.
Of the 2553 incident cases identified across all sites, 1181 were assessed in detail and form the case sample for case-control analyses. Across all sites, 1436 controls were recruited and assessed. This constitutes the largest and most extensive case-control study of psychosis ever conducted. The analyses reported here are based on the 900 cases and 1161 controls who completed at least part of the CECA and whose data had been entered into the programme database at the point of analysis.
Across all sites, there were broad similarities in the mean ages of cases at first contact (overall, 31 years) and controls (overall, 36 years) and in differences between them (i.e. in all sites, cases were younger than controls). Similarly, the proportions of men and women were similar. In all sites, for example, men comprised more than 50% of cases, reflecting what is known about a modest excess of psychosis in men compared with women. Again in line with previous research, there were more cases than controls from migrant and minority ethnic groups in all sites, irrespective of the overall proportion of migrant and minority groups in the general population. We adjusted for age, gender, and ethnicity in all analyses.
Our analyses of childhood adversities and case-control status produced four notable findings:
1) In combined samples including cases and controls from all sites, when presence or absence of seven forms of adversity were considered (household poverty, household discord, neglect, bullying, and psychological, physical, and sexual abuse), there was strong evidence of modest effects for all forms. That is, psychotic disorder (i.e. case status) was associated around a 1.5 to 2.0 increased odds of each type of adversity. These all held when adjusted for age, gender, and ethnicity.
2) There was some evidence that the effects were greater for more severe forms of adversity. For example, when only severe forms of physical abuse were considered, the odds ratios was higher than when only presence (inc. mild forms) or absence was considered (i.e. for physical abuse, OR 1.4 vs. OR 1.8).
3) The effects, despite some variation, were strikingly similar across the sites. This is illustrated in Tables 5 and 6 for severe physical and severe sexual abuse. With regard to physical abuse, for example, with the exception of Cambridge, all ORs indicate an increased odds of abuse in cases compared with controls, with most around 1.5 to 3.0. The same is roughly true for sexual abuse, though there is slightly more variability, perhaps due to sexual abuse being less common and there consequently being less precision in site specific estimates. In short, despite variations across sites in how common physical and sexual abuse are in controls, the odds of exposure were consistently around 1.5 to 3.0 times higher in cases.
4) There was strong evidence that the impact of adversities was cumulative. That is, the more adversities reported, the greater the odds of being a case (see Figure 1; score test for trend χ2 46.0 p < 0.001). This holds more specifically when physical and sexual abuse alone are considered (see Figure 2; score test for trend χ2 15.6 p < 0.001).
We found strong evidence that, in all sites, psychotic disorder was associated with a modest 1.5 to 2.0 times increased odds of each form of childhood adversity. These data provide strong evidence that previously reported associations with broader non-clinical psychosis phenotypes (e.g. low level psychotic experiences) extend to psychotic disorder. Further, there was some evidence that more severe experiences were associated with greater odds and that odds increased linearly with increasing number of adversities. These analyses are illustrative of the work ongoing in WP2 to investigate in detail hypotheses concerning environmental risk factors. Next steps include deepening these analyses by, for example, considered further the impact of timing and duration of exposures and extending to investigate interactions both with genetic risk and with other environmental risk factors.
EU-GEI WP2 collected extensive information on exposure to a full range of environmental factors, including adversities in childhood and exposure to recent life events in adulthood, in each site, from large samples of first episode cases and of population based controls. That the effects of exposures, such as childhood adversities (see Report 2), are broadly similar across sites allows us to pool data and to test more robustly hypotheses about interactions of environmental factors across the life course on odds (risk) of psychotic disorder. As an illustrative example, here we report preliminary analyses of examining synergistic (or combined) effects of childhood adversities and recent life events on psychotic disorder.
Of the 2553 incident cases identified across all sites, 1181 were assessed in detail and form the case sample for case-control analyses. Across all sites, 1436 controls were recruited and assessed. This constitutes the largest and most extensive case-control study of psychosis ever conducted. The basic demographic characteristics of cases and controls who completed the CECA are shown in Report 2, Table 3. The analyses reported here are based on the 784 cases and 905 controls who completed at least part of the CECA and whose data had been entered into the programme database at the point of analysis.
Analyses of the main effects of childhood adversities are presented in Report 2. Of specific relevance to the analyses in this report, there was strong evidence that the impact of adversities was linear, such that more adversities reported, the greater the odds of being a case (see Report 2, Figure 1). When modelled as an ordinal (continuous) variable, this was confirmed. That is, for every additional childhood adversity reported, the odds increased by around 20% (adjusted OR 1.23 95% CI 1.15-1.31 p < 0.001).
Recent life events
In line with findings on childhood adversities, there was evidence that number of recent events was associated in linear fashion with psychosis. That is, the greater the number of events, the greater the odds of being a case (see Figure 1; score test for trend χ2 30.2 p < 0.001). When modelled as an ordinal (continuous) variable, this was confirmed. Similar to childhood adversities, for ever additional life event reported, the odds increased by around 17% (OR 1.21 95% CI 1.09-1.26 p < 0.001).
Synergistic (combined) effects
When we considered the combined effect of childhood adversities and life events, we found some evidence of a modest – but statistically significant at p 0.05 – interaction on an additive scale. That is, the interaction contrast ratio of 0.04 (95% CI 0.01-0.08; p 0.025) indicates that for every additional childhood adversity and every additional event the odds ratio is 0.04 more than if there were no interaction.
(Note: The odds ratio and 95% confidence intervals of the product term for childhood adversities x life events show no evidence of interaction on a multiplicative scale.)
We found some evidence that childhood adversities and recent life events combine synergistically to modestly increase odds of psychosis, beyond the effects of each alone. These findings are broadly in line with a previous study of low level psychotic experiences and, more generally, with an emerging literature that suggests the effects of individual risk factors are amplified by the presence of other risk factors. There are some caveats. The effect found in these analyses is modest and the findings require confirmation in the full dataset. The analyses do, though, illustrate our approach to investigate environment x environment interactions and confirm that our sample size is sufficient to identify even relatively small interaction effects. Next steps include extending these analyses by examining other potential synergists effects of hypothesised causal partners (e.g. childhood adversities and cannabis use).
Risk Assessment chart for onset
An overarching goal of EU-GEI WP2, then, is to develop a socio-environmental risk assessment chart that comprises the major risk factors, is brief, and is predictive of onset of psychotic disorders. Our approach to developing a risk assessment chart comprises two steps:
1) Analyses of EU-GEI WP2 case-control data to quantify the main effects of hypothesised socio-environmental risk factors on psychotic disorder, as a basis for identifying items that will constitute the chart
2) Testing the predictive accuracy (power) of the chart
Here we briefly summarise analyses of main effects of hypothesised socio-environmental f actors found to be associated with psychosis (case status).
Of the 2553 incident cases identified across all sites, 1181 were assessed in detail and form the case sample for case-control analyses. Across all sites, 1436 controls were recruited and assessed. This constitutes the largest and most extensive case-control study of psychosis ever conducted.
Socio-environmental risk indicators and factors
Using the framework in Figure 1, Tables 2 and 3 show findings for those indicators and factors found to be associated with psychosis in initial case-control analyses (see also Reports 1-3) and which, therefore, partly form the basis for a socio-environmental risk chart.
Initial analyses (summarised in Table 1 and in Reports 1-3), then, suggest that aspects of the areas in which individuals live, indicators of position within and exclusion from the social structure, and negative experiences across the life course contribute to the risk of, or index risk for, psychotic disorder. Components of an initial risk assessment, derived from these analyses, are listed in Table 4.
The analyses presented here and in Reports 1-3 form the basis for an initial or first draft socio-environmental risk assessment chart. Further analyses will refine the items included and next steps will include testing the predictive accuracy (power) of the chart
WORK PACKAGE 3: DISCOVERY GENETICS
Genotyping and common variants
The investigators of WP3 have had numerous successes over the course of the EUGEI project, most notably their contribution to the work of the Psychiatric Genomic Consortium (PGC). The aims of this project were to gather all available genotype data for common genetic variants in schizophrenia and meta-analyse them together, to provide the most powerful genome-wide association study (GWAS) possible. EUGEI researchers have contributed samples and expertise to the PGC effort at multiple junctures, from the initial sample (n=21000) to the final dataset (35476 cases, 46839 controls).
This work has resulted in an increase in the number of genomic loci associated with schizophrenia at a genome-wide significant (GWS) level by an order of magnitude – from 15 in the first year of the EUGEI project to 128 in the final sample, published in Nature in 2014.
The EUGEI sample itself has been genotyped on the Institute of Psychological Medicine and Clinical Neurology (IPMCN) chip and has undergone quality control (QC). It now covers a total of 7722 samples that pass QC, from 20 centres across Europe. The sample has undergone preliminary association analysis using the mixed linear model method, which can take into account both the multi-national nature of the sample and the mix of family structures present. Meta-analysing the EUGEI association results with the PGC data finds a further 11 distinct GWS loci (Table 1).
Rare Single Nucleotide Variants
During the course of the project, WP3 investigators completed a large study of rare, exonic variation in schizophrenia based on exome chip technology (5585 cases and 8103 controls). Consistent with what are now the expectations for a polygenic disorder, no individual rare variant was associated at GWS levels, the most strongly associated variant being within the gene DENND2D (p=1x10-5). However, supporting our hypothesis that true rare variant risk alleles are present on the array, we identified an excess of rare variant associations in genes mapping to loci that we and others previously implicated in schizophrenia.
We also found association at one gene, WDR88, at the gene-wide equivalence of genome-wide significance. This analysis has also extended the evidence (based upon different classes of mutations in other datasets) for the involvement of a set of genes comprising FMRP targets (p=0.0048).
As the IPMCN chip includes these rare, exonic variants, we were also able to extend our analyses to the EUGEI sample. To confirm that true signal is captured by this analysis, we undertook polygenic score analysis based on rare variants, which confirmed that risk alleles developed from one sample are overrepresented in another (p=7.5x10-5 using SNVs with MAF < 1% and discovery p<0.5). Remarkably, the most significant single variant in our earlier exome chip study was also highly significant in the EUGEI data (DENND2D), the results giving a combined meta-analysis p-value that reaches genome-wide significance (p=3.95x10-8).
Copy number variants
Another major success of WP3 has been the study of copy number variants (CNVs), especially de novo CNVs. These are recently arisen CNVs that are present in an individual but not their parents. They have not been subject to any selective pressures, and so may have a greater effect on the phenotype of an individual than inherited CNVs.
During reporting period 2, WP3 researchers compared 662 schizophrenia proband-parent trios to 2623 control trios. This work showed that schizophrenia cases suffer from de novo CNVs at a rate 2-3 times higher than the general population (Table 2; Table 3). CNVs have now been called in the EUGEI data, and this finding was further borne out by CNV analysis of the EUGEI proband-parent trios. 27 de novo CNVs in 702 schizophrenia trios (3.8%) and 3 de novo CNVs in 186 control trios (1.6%) were found – a twofold enrichment.
In addition to the de novo work, WP3 members undertook an analysis of a new dataset comprising 6882 cases and 11255 controls and merged this with the extant schizophrenia world literature (case sample sizes for meta-analysis ranging from 12,000-21,200 cases varying by locus). In this, the largest systematic CNV analysis of schizophrenia, we found 11 CNV loci now meet criteria for genome wide significance. These included for the first time a locus on 16p13.13 and an imprinted maternally transmitted duplication at the Angelman/Prader-Willi Syndrome locus; for both of which earlier work had provided only modest evidence.
Seeking further novel CNVs, WP3 investigators undertook a detailed systematic (i.e. data not derived from the literature on specific CNVs), gene focused CNV study with discovery and follow up in 21,000 cases and 25,000 controls. This highly powered analysis has not identified additional CNVs at levels close to genome wide significance. Overall, this, the world’s largest analyses of schizophrenia CNVs, suggests the CNVs that can be confidently considered schizophrenia related are carried by about 2.4% of cases and about 0.6% of controls, with estimated effect sizes (OR) ranging from 2-60. If there are (as seems likely) additional CNVs to be discovered, they are too small to be detected by GWAS technology, or their individual contributions to overall risk of disorder is less than those discovered to date, resulting in insufficient power of the current analyses.
WP3 investigators evaluated the rate of known schizophrenia CNVs in all analysed EUGEI samples for which we have phenotype data (Table 4). All previously identified schizophrenia risk CNVs have higher frequencies in cases than controls, providing further support for their role in the disorder, albeit at slightly reduced frequency overall (1.6% v expected 2.5% in cases). This reduced frequency may reflect the inclusion of incident as well as prevalence cases, and consistent with this, WP6 (prevalence cases; 2.1%) has a higher rate than WP2 (incident cases; 1.29%). It may also reflect a broader non-affective psychosis rather than schizophrenia phenotype among EUGEI data. Overall, schizophrenia risk CNVs were highly enriched among schizophrenia cases compared with controls/siblings (P = 0.00038).
Multiple novel loci containing CNVs of possible relevance to schizophrenia aetiology have been identified by WP3 members. These include risk-increasing deletions and duplications across the genome, and also interestingly a protective duplication at 22q11.2. This is a particularly important finding as identification of the dosage-sensitive gene(s) at 22q11.2 might suggest pharmacological intervention that offers protection from schizophrenia in individuals who are otherwise at high risk.
WP3 researchers have performed pathway analysis on both rare and common variants, using candidate pathways chosen on the basis of experimental data. In reporting period 2, the de novo CNV data produced a striking result. In comparison to several different reference CNV control datasets, case de novo CNVs were highly significantly enriched for post-synaptic proteins and in particular members of the N-methyl-D-aspartate receptor (NMDAR) and neuronal activity-regulated cytoskeleton-associated protein (ARC) complexes.
In reporting period 4, these categories were further supported by exome sequencing data and further CNV trio work. The fragile X mental retardation protein (FMRP1) target and actin filament bundle assembly gene sets also emerged as new findings from the sequencing data, while the case-control CNV implicated the GABAergic post synaptic system in schizophrenia (P=0.0004).
During reporting period 5, WP3 investigators have extended and finalized these analyses in a larger CNV study (N cases = 11355; N controls = 16416). The findings confirm the NMDAR network (Odds ratio (OR) = 3; P<10-8), ARC complex (OR = 1.7; P=0.0003) and provide further nominal significant support for FMRP targets and calcium channels. Our novel finding of the involvement of GABAergic genes is also extended (OR = 2.51; P = 3x10-6).
WP3 investigators also found that large schizophrenia deletions are significantly enriched in a hypoxia set (p=2.5x10-5). This result was not driven only by known schizophrenia deletions, as it remained significantly enriched even when previously found schizophrenia CNVs were removed (p=5.4x10-3).
Pathway analysis was also performed in the study of rare, exonic variation, using the method SKAT-O and a set of candidate pathways. A pathway consisting of genes from GWS associated regions in the PGC study of common variation showed significant association (p=0.02) showing there is some overlap of signal between rare and common variation. Genes targeted by FMRP1 also showed significant schizophrenia association (p=0.0029).
To perform pathway analysis of common variation, WP3 researcher Holmans developed the novel method ALIGATOR, and more recently an approach in which the results of multiple pathway analysis methods are synthesised (ALIGATOR, INRICH, JAG, MAGENTA, FORGE and SETSCREEN). When applied to the combined GWAS datasets, the candidates derived from rare variant analyses (FMRP1 targets, calcium channels, NMDAR) were significantly associated (P= 1x10-5 - P= 3x10-4). The convergence of results from CNVs, common variation, and rare point mutations onto these three pathways is striking, and provides the strongest evidence yet that they have a role in the aetiology of schizophrenia. We also performed pathway analysis using a wider set of functional categories taken from the GO, BioCarta, KEGG, REACTOME, PAN-PW, NCI and MGI sets. Notably, the most significant result was the KEGG dopaminergic synapse category, giving support to the dopamine hypothesis of schizophrenia (P=1x10-7 ; Table 5).
To summarise, WP3 has made considerable progress in the study of schizophrenia across the five years of the EUGEI project. The EUGEI sample has been genotyped (see Table 6 for sample numbers) and initial association analyses completed. Contribution of EUGEI genotype data to the PGC meta-analysis study has helped increase the number of genetic loci robustly associated with the disorder by an order of magnitude. CNV work has demonstrated an excess of CNVs previously linked with schizophrenia in the EUGEI sample and other samples, as well as an excess of de novo CNVs in schizophrenia cases. It has also identified several novel CNVs of potential relevance to schizophrenia aetiology. Lastly, pathway analysis of data from multiple sources is converging on several functional categories, most notably the NMDAR, calcium channel and FMRP1 target gene sets.
WORK PACKAGE 4: EXPERIMENTAL GxE
Neural mechanisms underlying environmental risk factors
Prior epidemiological studies investigating the risk factors for schizophrenia are strongly suggestive of a complex interaction between genetic susceptibility and multiple environmental risk factors. Specifically, urban upbringing, migration, low IQ, low socioeconomic status (SES) and the use of illicit drugs are among the best established environmental risk factors associated with the disease.
Study-1: Developing a social stress paradigm for fMRI environment
In order to investigate neural mechanisms underlying environmental risk factors, we developed a flexible imaging paradigm to elicit an acute stress response in magnetic resonance imaging (MRI) setting, comparable to previously established paradigms for stress induction outside the scanner. We use this paradigm to further analyze the possible links between well-established environmental risk factors such as urban upbringing and minority status; and to measure psychological and neural responses to acute stress in these high risk groups. Following the frequently used and efficient approach of the Trier Social Stress Test (TSST) we present a live video stream of a panel to the participant during the scanning procedure. The panel is monitoring and giving simultaneous visual feedback with a “buzzer”, as well as verbal feedback which are given between the two scan sequences in order to induce social-evaluative threat. Continuous time pressure during this cognitively demanding task, as well as adaptive speed and difficulty of the tasks resulted in the uncontrollability. We have successfully developed a flexible social stress paradigm and implemented it in Study-2 and Study-3. The results of social stress paradigm are already published in high-impact peer-reviewed journals.
City living and urban upbringing affect neural social stress processing in humans
Lederbogen, F., et al., Nature, 2011. 474(7352): p. 498-501.
Neuroimaging evidence for a role of neural social stress processing in ethnic minority-associated environmental risk
Akdeniz, C., et al., JAMA Psychiatry, 2014. 71(6): p. 672-80.
Study-2: Urban upbringing and neural response to stress
Using our social stress paradigm we have investigated the effects of urbanicity on brain function during acute stress. Urbanicity was quantified as follows: category 3: city with more than 100,000 inhabitants; category 2: town with more than 10,000 inhabitants, and category 1: rural area. For urban upbringing, these numbers were multiplied by the number of years living in the area up to age fifteen and added. Thirty-two healthy participants with rural as well as urban upbringing were studied. Functional MRI (fMRI) was performed on a 3 Tesla Siemens Magnetom Tim Trio scanner. Current urban living was associated with amygdala activity during stress, which increased stepwise from subjects living in the country to those living in small towns, and was highest in city dwellers. In contrast, urban upbringing during first 15 years of life was associated with differential activity in the perigenual ACC (pACC), increasing linearly with highest activation in participants entirely brought up in cities. We have, therefore, showed a dose-response relationship between urban upbringing and currently living in an urban environment and neural response to stress.
Study-3: Altered neural stress processing in migrants
Risk for the severe brain disorder, schizophrenia, is at least doubled in first- and second-generation migrants. While this has been widely attributed to the social environment, the underlying neural mechanisms were unknown. As summarized above in Study-2, we previously found that the perigenual cingulate cortex (pACC), a key region for regulation of limbic activity, negative affect and stress, was selectively associated with urban upbringing, another established environmental risk factor for schizophrenia. We therefore hypothesized that migration status might impact on the same neural system. We studied a sample of 80 healthy native German participants, half of which were second-generation migrants and half non-migrant controls. Groups were carefully matched for a broad range of sociodemographic characteristics including urban exposure. We examined second-generation migrants, a group with schizophrenia risk equal to that of first-generation migrants, to eliminate variance from pre- and perimigratory effects, and to control for possible environmental exposures unrelated to immigration status, but differing between the country of origin and Germany. During social stress condition, migrants exhibited significantly increased activity in pACC compared to non-migrants, but no other brain region. This confirmed our hypothesis and provides the first direct evidence for altered neural social stress processing in migrants. We additionally tested the effects of perceived group discrimination on stress-related activation in our migrant sample. Perceived group discrimination was significantly correlated with pACC activation, as well as ventral striatum.
Study-4: Nonlinear cumulative risk effects of migration status and sex on perigenual anterior cingulate volume
We studied 140 healthy young adults (70 native German participants and 70 individuals with migration background). Groups were equal in mean age, as well as gender, urban exposure and school education. Magnetic resonance imaging (MRI) was performed on a 3 Tesla Siemens Magnetom Tim Trio scanner. The effect of migration background on gray matter (GM) volume was tested in a multiple regression analysis with group as covariate of interest. In addition, since male sex impacts brain structure and has been associated with increased schizophrenia risk, earlier age of onset, and course of illness , adverse interactions with migration background was also in focus. Thus, potential interaction effects of the variables on gray matter volume were tested in an ANOVA model with sex as a factor, and a sex by group interaction term as covariate of interest; age, education, and urbanicity as nuisance covariates. We specifically focused on pACC as a region of interest (ROI), since pACC is a brain area where environmental risk factors seem to converge. There was no significant main effect of groups in pACC. We tested for interaction effects of group (migrants vs. Germans) by sex on gray matter volume. We detected a significant group by sex interaction effect in pACC. Post-hoc analyses reveal a significant difference in pACC GM volume between males and females in migrants, but not in Germans. In conclusion; our main finding was consisted of an inverse relationship between migration background, sex and gray matter volume in the pACC; indicating significantly reduced pACC gray matter in males with migration background.
Publications are in preparation.
Study-5: ACC gray matter volume correlates with social mobility
We studied 140 healthy young adults (76 migrants and 64 Germans). Perceived current social status was measured using the McArthur rank-order Subjective Social Status Scale. Perceived social status of the birth family environment was assessed with an adapted version of this measure in which participants ranked the perceived social status of their own parents at the time of their birth. We then calculated individual “social status mobility“ scores which are reflecting the difference between perceived current and birth family social status. Chronic stress was measured with the Chronic Stress Screening Scale. MRI was performed at 3T using a high-resolution T1-weighted sequence. Images were processed using the SPM8 VBM toolbox and analyzed using general linear models with random-effects group statistics. Post-hoc mediation analysis was performed using PROCESS. We observed no significant main effect of perceived current social standing on pACC volume in either group, but a significant group-dependent interaction suggesting a positive association in Germans and a negative association in the ethnic minority group. In addition, we detected a significant group-dependent interaction of the association of social mobility scores and pACC volume, with no association of the measures in the German majority group but a highly significant negative association in the migrant group. Posthoc mediation analysis showed an indirect effect of social status mobility on pACC volume through chronic stress (a*b = -0.0018 CI(95) = [-0.0042; -0.0002]).
Publications are in preparation.
The studies have shown a convergent impact of validated environmental risk factors for schizophrenia on anterior cingulate cortex (ACC), a region where structural and functional changes have been identified in patients already at the first episode of psychosis. Biological evidence has been provided showing that social stress mediates the effects of environmental risk factors for schizophrenia; and possible accumulation of several risk factors increase the risk which demonstrates itself as functional and/or structural changes in neural processing and architecture of ACC. Lastly, we have demonstrated the significant potential of investigating associations from psychiatric epidemiology using neuroimaging, and we encourage future research on the neural convergence mechanisms of genetic and environmental risk factors in this circuitry.
This study aimed to assess aspects of salience attribution and its relation to subclinical psychosis, as measured by the Community Assessment of Psychic Experiences (CAPE). Participants were sampled from the East Flanders Prospective Twin Survey (EFPTS), focusing on individuals aged beween 15 and 35 years of age. The EFPTS is a prospective, population-based registry of multiple births in the province of East Flanders, Belgium. The following experimental tasks were designed to assess aspects of salience attribution: psycho-babble task, white noise task and signal detection task and reversal with reward.
We were able to recruit 398 twin pairs (796 individuals; 59% females). These individuals had mean positive CAPE scores of 0.61 (SD 0.38) and negative CAPE scores of 1.68 (SD 0.35).
Preliminary analyses show that reward sensitivity weakly predicted the Negative dimensions of the CAPE-42 (r=-.12). Strength of reward reversal was not associated with any dimensions of psychotic experiences. The Psychobabble Task and the White Noise Task in preliminary analyses also seem to correlate with the Positive (psychotic) dimensions of the CAPE.
The different salience tasks that were tested seem promising as they correlate with subclinical dimensions of the positive and negative symptom scales relevant to schizophrenia. Future work will be able to provide more in-depth analyses investigating risk factors associated with salience misattribution as well as risk-modifying variables (e.g. vulnerability genes) in the developmental pathway towards psychosis.
After the substantial delay in trying to get access to specific radioligands for human use to detect the availability of dopamine D2- and cannabinoid CB1-receptors in healthy volunteers under the placebo-controlled influence of either Δ9-tetrahydrocannabinol (Dronabinol) or synthetic cannabidiol or a combined administration of both, we finally developed a contingency plan replacing the part of the study where there molecular imaging of cannabinoid CB1-receptors by use of 18F-MK-9470 was planned to a design that investigated neural synchrony and event related potential in a 128-channel EEG recording. We have successfully performed the fMRI/EEG part of the study. 61 volunteers (including one drop-out) with appropriate COMT genotype have been identified, enrolled and randomized within the trial.
This trial has provided a plethora of data that impact our understanding of the effects of different compounds of the cannabis plant and their interaction substantially. Even the early results have been surprising to some extent. While clinically not very different from the effects of ∆9-THC (dronabinol) alone, the combined administration revealed a different pattern of brain activation that resembles those of antipsychotic compounds. The study underlines that cannabidiol itself is very well tolerable with only few if any clinical effects in healthy individuals. These findings may impact legislative approaches in the current discussion of legalizing cannabis products in a number of EU member states. A more detailed analysis is necessary to differentiate pharmacokinetic from pharmacodynamics effects. If the combined administration of cannabidiol and ∆9-THC (dronabinol) will prove as superior in terms of pro-psychotic effects of ∆9-THC (dronabinol) alone, this may influence the type of cannabis derived material that may be legalized for recreational use (e.g. ∆9-THC (dronabinol) enriched material without any cannabidiol may still be banned). Beside its societal implications, this study widens our understanding of the mechanistic actions of cannabidiol, which is under development as an antipsychotic without support from major pharmaceutical companies but with funding from the EC. Our preliminary findings suggest that the data set derived from the final analysis will support a Go/NoGo decision in several directions of future drug development for cannabidiol.
Interestingly, a first, preliminary analysis indicated that an interaction of the observed effects with the COMT genotype was not present in the study. Additional data analysis on relevant genetic factors has been performed in collaboration with P1. Given the negative finding on COMT and the omission of some of the second part of the study due to unforeseen technical problems, WP4 was fortunately in the position to anticipate to recent scientific developments suggesting that the mechanisms of cannabis exposure are shared with those of hypoxia exposure. WP4 therefore took the opportunity to analyze genetic effects of exposure in a unique, deeply-phenotyped cohort with available prospective data on exposure to hypoxia, that became available to the WP in collaboration by P1/MUMC. Data collection in this cohort was already performed. WP4 genotyped the DNA and statistical analysed the data of this cohort, such that the project between P4/CIMH and P1/MUMC could be performed in the limited time left in the project. The results of the study identified distinct genetic variants in AKT1, BDNF, CHRNA7, GABRB2, PLXNA2, RELN, RGS4 and YWHAE that moderated the impact of exposure to hypoxia on the expression of psychotic symptoms, and therefore provide important knowledge for candidate genes for further mechanistic studies on cannabis.
WORK PACKAGE 5: GxE PRODROME
Onset of schizophrenia
The consensus in the field of Clinical High Risk or ‘Ultra High Risk’ (UHR) is to focus on the onset of psychosis rather than onset of schizophrenia per se. This is for two main reasons, 1) a number of UHR individuals develop non-schizophrenia psychosis such as bipolar disorder and psychotic depression (Fusar-Poli et al., 2013), and 2) the formal diagnosis of schizophrenia requires 6 months of continuous disturbance [DSM-V criteria (American Psychiatric Association, 1994)] and the required follow-up period extends beyond the end of the EU-GEI project. Thus in the current report we examine the onset of psychosis in individuals recruited in WP5 of the EU-GEI project.
To the authors’ knowledge this is the first time that that the trajectory of transition to psychosis from UHR has been monitored prospectively in a large, globally representative sample. We have compared the percentage of UHR individuals who transition to psychosis, and the temporal course of transitions in EU-GEI WP5 to two meta-analyses based on previous individual UHR studies (Fusar-Poli et al., 2012; Kempton et al., 2015).
Percentages of UHR individuals who have transitioned to psychosis
358 UHR individuals participated in EU-GEI WP5 and as of March 2015, 50 participants transitioned to psychosis (14% of the sample).
The fraction of participants who transitioned to psychosis varied from 26% in Basel to 5% in Barcelona (table 1, figure 1). However, a chi squared test across the 11 centres did not detect a significant effect of centre on transition rate (χ²(10)=12.4 p=0.25).
Temporal course of transitions
In the analysis of 31 individuals where the transition date was available we produced a plot showing the transitions over time up to 2 years (=730 days) since enrollment (figure 2). This preliminary transition data was reformatted to the percentage of transitions given that transition occurs at 2 years. Thus the curve is normalized with 100% of transitions occurring at 2 years (blue line, figure 3). The normalized transition curve was compared to a corresponding regression curve published in an independent meta-analysis (red line, figure 3). The preliminary EU-GEI data was a close match the regression curve from the meta-analysis (figure 3, R=0.99 p<0.001).
The overall transition rate to psychosis in the EU-GEI WP5 sample was 14%. This is lower than the value of 29% at 2 years reported in the meta-analysis by Fusar-Poli et al (2012) however it has been recognized that transition rates have been falling over the last decade in UHR studies (Yung et al., 2007) and the 29% figure quoted by Fusar-Poli et al (2012) included earlier studies with high transition rates. Thus the 14% may be more representative of current transition rates. Although we reported nominally different transition rates in the different centres there was no significant effect of site on the percentage of those who transitioned to psychosis. We were able to demonstrate that although preliminary, the temporal course of transitions in EU-GEI matched previous findings very closely. Future research on this comprehensive dataset and our other WP5 reports will investigate which particular factors may predict transition to psychosis in UHR individuals.
Risk Assessment chart for onset
The aim of the chart is to use machine learning techniques applied to clinical, environmental, cognitive and neuroimaging measures in the high risk phase to predict which individuals will go on to become psychotic. This would allow clinicians to target preventative interventions in those most at need.
Clinical, Environmental and Cognitive Measures
Using clinical, environmental and cognitive data, SVM was able to discriminate between UHR who subsequently made transition to psychosis and UHR who did not with an accuracy of 60.9% (specificity: 62.5%; sensitivity: 59.4%), which was statistically significant (p = 0.044). Figure 2 indicates the weight vectors which contributed towards the classification of UHR-T vs UHR-NT.
Using gray matter volume information, SVM was not able to discriminate between UHR who subsequently made transition to psychosis and UHR who did not with a balanced accuracy of 50% (specificity: 66.7%; sensitivity: 33.3%), which was not statistically significant (p = 0.47).
Combining both gray matter information and the clinical, environmental and cognitive measures, SVM could not discriminate between UHR-T and UHR-NT individuals who subsequently made transition to psychosis and UHR who did not with an accuracy of 50% (specificity: 53.3%; sensitivity: 46.7%), which was not statistically significant (p = 0.54).
The current analysis represents the first time that a machine learning algorithm has been applied to a prospectively collected global UHR dataset to predict transition to psychosis. Clinical, environmental and cognitive data collected at baseline was able to predict subsequent transition to psychosis with a modest level of accuracy. In our preliminary SVM analysis of structural MRI data we were not able to predict transition in UHR. Integrating MRI and the non-imaging data did not lead to improved classification accuracy. The analysis of data is at an early stage, however we have been able to successfully demonstrate the feasibility of collecting and analysing standardized clinical, environmental, cognitive and neuroimaging data across 11 international centres. In the current analysis we have chosen a limited number of variables but we will be aiming to expand this in future. In particular we were not able to include genetic data such as polygenetic risk score because of delays in completing the GWAS analysis. However including this information with other data may increase the algorithms’ predictive accuracy. A limitation of the current methodology is the requirement of including equally sized groups, however in UHR research the transition group is inherently smaller than the non-transition group. In both analyses it was necessary remove UHR-NT data to balance the group size, which reduces statistical power, thus we will also be investigating other machine methodologies which will enable us to include a larger number of participants. By improving both the methodology and fine-tuning the input variables we look forward to substantially improving the accuracy of the risk assessment tool.
Momentary Assessment Tool for onset
We successfully completed data collection using our newly developed, electronic momentary assessment technology device (PsyMate®) and met our initial targets for recruitment and ESM assessment of UHR subjects. In total, 112 subjects (Amsterdam & The Hague, n=49; London, n=45; Melbourne, n=18) were assessed with this new technology, providing a total of 3393 observations. Of these, we identified 81 subjects (72.3% of 112; Amsterdam & The Hague, n=26; London, n=40; Melbourne, n=15) with valid ESM assessments (i.e. ≥20 observations over the 6-day assessment period), yielding a total of 3072 valid observations measured with the PsyMate® at baseline. This provided good evidence that the use of this novel technology is feasible in UHR populations. Further, we identified no risks associated with using the PsyMate® throughout the study period.
1) We found strong evidence of elevated stress sensitivity in UHR subjects, indexed by strong negative emotional reactions in response to event-related (B=0.41 95% CI 0.29-0.53 P<0.001) activity-related (B=0.29 95% CI 0.25-0.33 p<0.001) and social (B=0.13 95% CI 0.11-0.15 p<0.001) stress in daily life.
2) Elevated stress sensitivity was, in turn, associated with a three- to four-fold increased odds of developing intense psychotic experiences in daily life (event-related stress sensitivity, OR 3.48 95% CI 1.44-8.46 p=0.006; activity-related stress sensitivity, OR 3.94 95% CI 1.85-8.42 p<0.001; social stress sensitivity, OR 3.98 95% CI 1.86-8.52 p<0.001).
3) At baseline, stress sensitivity was markedly elevated in UHR subjects who transitioned to psychotic disorder during the follow-up period (n=6, with 218 valid ESM observations at baseline), as evidenced by significantly stronger emotional reactions in response to event-related stress (B=0.32 95% CI 0.24-0.41 P<0.001) compared to those who did not transition to psychosis (B=0.14 95% CI 0.12-0.16 P<0.001; likelihood ratio teststatus × stress, χ2=15.74 p<0.001; see Figure 1).
4) Event-related stress sensitivity was significantly elevated in UHR subjects with more severe symptoms and lower global functioning as measured with the GAF at 1-year follow-up (B= 0.19 95% CI 0.13-0.24 p<0.001) (compared to those with less severe symptoms and higher global functioning, B= 0.10 95% CI 0.01-0.19 p<0.001; see Figure 2).
WORK PACKAGE 6: GxE VULNERABILITY & SEVERITY
There has been substantial progress in understanding of the interplay between genetic and environmental vulnerability factors underlying pathoetiology of schizophrenia. A growing body of evidence suggests exposure to environmental factors such as cannabis use, childhood trauma, and urbanicity influence susceptibility to psychosis across general population and individuals at risk. However, research on gene-environment interaction (G x E) in psychosis has been mostly limited to studies using candidate gene targeting in small samples that tend to yield results that are not replicable. Therefore, there is a certain need for large-scale investigations of the impact of G x E on psychosis liability and severity, and WP6 of EU-GEI is aimed to fill this gap in the field.
To assess phenotypes associated with vulnerability (neurocognition, social cognition, psychosis proneness, and cerebral structural measures) and severity (neurocognition and cerebral structural measures), a prevalence sample of patients with schizophrenia, their siblings, and ethnicity-matched controls were recruited and clinically, environmentally, and genetically assessed in 8 sites: Turkey (the National Schizophrenia Network [partners AU and Omega Pro]), Spain (the CIBERSAM network [partners SERMAS and SERMES]), the Netherlands (as part of Dutch GROUP study [partners MUMC and AMC]), and Germany (partner LMU).
Samples of 1000 patients, 1000 of their sibs, 700 of their parents, and 1000 ethnicity-matched controls in the Netherlands; and 1000 patients and 2500 ethnicity-matched controls in Germany were already collected and available at the beginning of the project.
Samples of 500 patients, 500 of their siblings, and 500 ethnicity-matched controls in Spain; and 1000 patients, 1000 of their siblings, and 1000 ethnicity-matched controls in Turkey were aimed to be collected. Both sites obtained approval of the local Ethics Committee before commencing the enrolment. To collect information about phenotypes, instruments to assess phenotypes were translated and back-translated into their original language to improve accuracy. Extensive training on instruments and interview skills were provided to research assistants to ensure data quality. Initial assessments were reviewed, and possible difficulties were anticipated. In addition to the EU-GEI web-based training designed to control and increase inter-rater reliability, regular meetings were held to discuss case vignettes in both Spain and Turkey. Site visits were held in order to evaluate and standardize interviews. Missing data patterns were examined and strategies were updated in order to minimize missing data. These strategies were discussed with research assistants. To collect information about genotype, blood samples were obtained, extracted, and shipped to the UK (partner CU).
As reported in the progress reports of WP6, recruitment rates were below the original plans, both in Spain and Turkey. Therefore, a risk contingency plan was developed and implemented during the project. As part of the risk-contingency plan to ensure targets for recruitment are achieved, two sites were added (Belgrade and Verona) at no extra cost., while partners AU and Omega Pro prolonged their recruitment periods and continued to recruit participants until the end of RP5, at no extra cost.
Combined with the samples available in project year 1, WP6 has recruited and clinically, environmentally, and genetically assessed an overall of 3,227 cases, 2,294 siblings, and 5,001 controls. Genetic samples were acquired from all participants. At the end of RP5, WP6 targets for cases, siblings, and controls were reached for 92.2%, 91.7%, and 100%, respectively (see Table 1 for detailed recruitment numbers at each site). All collected data were entered, and genetic samples were shipped to the UK (partner CU).
Data from a WP6 subsample of 300 patients, 300 sibs and 300 controls has already been analysed. Using structural MRI, and (i) cross-sibling within-trait, (ii) cross-sibling cross-trait and (iii) within-sibling, cross-trait correlations of cognition, WP6 researchers have quantified psychosis proneness and MRI phenotypes,and the impact of environmental factors on these correlations as indirect evidence for G x E was investigated. These WP6 investigations revealed e.g. G x E interplay between FKBP5 and childhood trauma to increase psychosis expression, correlations between stress-sensitivity and pituitary volumes, microstructural abnormalities and familial risk, and reduced cortical thickness as an outcome of differential sensitivity to environmental risks in schizophrenia (see references below).
Work Package 7: GxE Course
WP7 has successfully completed the recruitments and assessments of 1120 patients, 1057 siblings, 919 parents and 590 healthy controls at baseline. Patients, siblings and controls were furthermore assessed at two later time points during a period of five years in order to address determinants of early course. Data collection has been completed, genotyping has been performed and data quality control has been performed for almost all variables. During the project period, WP7 has successfully investigated genetic, environmental and clinical variables in the GROUP cohort and has published a first wave of the findings of the baseline measurement in several high-impact research manuscripts; see below for a selection of these manuscripts. Data on early course of schizophrenia is currently under analysis, making use of the statistical tools developed within WP8 in EU-GEI.
In the course work package, we established that specific genetic variation may interact with childhood adversity and cannabis use in predicting onset and persistence of psychosis. We found that familial risk for schizophrenia is associated with a 10-20-fold increased sensitivity to the psychotogenic effects of cannabis, and examined underlying mediation by differences in craving. We furthermore established patterns of co-segregation of neurocognitive traits, personality factors and symptom measures in psychotic disorders, and how these impact course of illness. One genetic variant was found to predict covariation of psychosis and neurocognitive impairment. We furthermore found that gene-environmental interplay appears to be the rule rather than the exception in psychotic illness.
User-led stigma project: D-STIGMI (Destigmatizing Mental Illness): a randomized controlled trial
Stigmatisation is an important phenomenon with negative consequences that many people with (a diagnosis of) schizophrenia or other psychotic disorder will have to deal with. In the project, possible effects of stigmatisation and its consequences on the course and outcome of schizophrenia have been investigated. Different types of stigma - public stigma, self-stigma and label avoidance - may each have detrimental effects on this disorder. For the purpose of the project, stigmatisation is thought of as an environmental risk factor that exerts its influence along diverse pathways. These pathways have been investigated in a randomized controlled trial, using qualitative (focus groups) and quantitative (interviews) instruments. The Experience Sampling Method (ESM) was used as an outcome method to assess stigma-related experiences in daily life. The ESM holds several advantages over traditional cross-sectional assessment strategies (questionnaires or clinical interviews). It has allowed reporting temporal relations between variables (what happened and how did you feel before, during and after the event). It has high ecological validity (reports actual emotions in normal situations in situ) and has yielded detailed information on subjective experiences and engagement.
The research proposal was written by an expert by experience and service users were involved in diverse elements of the research process, such as data collection by being involved as trainers of the psycho-education coping skills training (intervention of the RCT) and the newspaper reading group (control condition of the RCT). They have also been included in the discussion of findings.
The trial has shown that both group programs have positive effects for people with severe mental illness (SMI). We can further conclude that ESM can be used to measure stigma-related variables in daily life of people with SMI. As part of the dissemination activities the results of on the effectiveness of group programs, based on ESM-results, in improving the individual’s quality of life has been (and will be) presented on at conferences. An article is in progress.
After completion of the trial a tool has been developed on how to organize a user-developed intervention to reduce the effects of stigma. The tool is available via the project’s website.
Momentary Assessment Tool for early course
Mental disorders such as depression, psychosis and anxiety are influenced by such a complex interplay of factors that it is extremely difficult to develop accurate predictive models. Complex dynamical system theory may provide a new route to assessment of personalized risk for transitions in depression. In complex systems, it has been demonstrated that generic early warning signals (EWS), signaling critical slowing down of the system, precede critical transitions. In the EU-GEI project, we tested the hypothesis that transitions in mental states such as mood, as assessed with PsyMate experience sampling technology, may follow similar principles as reported in other complex dynamic systems such as climate systems or eco-systems.
In order to show proof of principle, we conducted an experiment –double-blind with regard to timing of antidepressant dose-reduction– to examine whether critical transitions in mood are preceded by critical slowing down in mood fluctuations. Experience Sampling Methodology (ESM) was employed over a period of 239 days and the SCL-90-R depression subscale was administered weekly.
It was found that all hypothesized EWS were present and preceded a significant and clinically relevant shift in symptoms. Furthermore, we demonstrated the existence of this phenomenon prospectively within data of a single individual. Critical slowing down in mood fluctuations preceded a shift in the mood system. This confirms the idea that transitions in mental symptoms follow principles of complex dynamical system theory, and suggests a novel way to construct personalized warning systems for onset and relapse to depressive, psychotic and anxiety episodes.
The PsyMate prediction tool
These EU-GEI findings support the hypothesis that mental states like depression and psychosis can be characterized as a complex dynamical system: generic early warning signals, which predict critical transitions in many different dynamical systems (Scheffer et al., 2009, Scheffer et al., 2012, van Nes and Scheffer, 2004), were present in self-reported measures of momentary affect. Following an antidepressant tapering, we observed a clinically relevant transition to depression, which was preceded by CSD. This suggests that the same warning signals that indicate upcoming changes in financial markets, the climate, water quality in lakes, and the onset of epileptic insults, may also apply to clinically relevant shifts in psychopathology.
This EU-GEI report on PsyMate technology examined intra-individual changes over time during critical mood shifts, utilizing momentary assessments methodology over a period of several months. The findings provide support for the hypothesis that inertia – the lingering of mental states over time – is associated with an increased risk for psychopathology (Koval et al., 2012, Kuppens et al., 2012) and that PsyMate technology can be used to predict critical transitions. Feedback loops involving strong connections between such mental states create negative spirals that may pull individuals into a vicious circle of emotional experience from which it is difficult to escape. Eventually, this may induce a so-called catastrophic shift in emotional experience, such that the individual enters the alternative stable state of psychopathology. This theory accords well with the changes in network structure that we observed in the current study. Both the number of significant connections and the strength of these connections appear to increase as time unfolds. At the start of the experiment, depressive symptoms were absent, and the network featured a low level of connectivity, suggesting it is at rest. However, when antidepressant medication is tapered off, the structure of the network changes: mental states begin to start lingering, and to show increased effects on each other. Importantly, and unexpectedly, the changes in network structure continue to intensify also after the transition took place. This may be an indication that yet another critical transition was imminent; it is impossible to test this hypothesis as the patient resumed medication. Alternatively, the results may indicate that depression is both caused and sustained by increased connectivity of symptom networks. Further research is needed to unravel this issue. Also, future studies should try to replicate the current findings in multiple individuals. These should individuals with and without a critical transition in symptoms, in order to confirm that EWS are present in those with a transition and are not present in individuals without a transition.
The generic principles of complex dynamical systems as indexed by PsyMate technology in EU-GEI may offer new possibilities for explaining and predicting shifts in early psychosis and depression, even if we cannot yet fully grasp the mechanistic and etiological details of the condition. This approach deviates from traditional scientific approaches in that stability of the system, based on present EWS, is assessed per system. For applications in the field of psychiatry this means that estimation of vulnerability for transition, using PsyMate technology, does not require averaging over a group of people, as is the default approach of analyzing data, but can be estimated per individual. The PsyMate approach may offer a new way in the field of psychiatry to acquire personalized and clinically relevant information for prediction of important transitions.
WORK PACKAGE 8: GxE DATA & STATISTICS
New methods of GxE analysis and modules for PLINK have been developed. As single SNP analysis may lack statistical power, gene based and set based tests have been proposed to combine evidence from multiple SNPs as well as genetic main and interaction effects. WP8 researchers have extended gene-based test for disease association (GATES) (Li et al., 2011), into a gene-based test for association with multiple phenotypes (mGATES) (Van der Sluis et al., 2015). This found many novel trait associating loci in a metabolomic dataset. A further spin-off from this is the development of SPS, a power calculation tool for gene set based tests, on the basis of simulating phenotypes for real genotype datasets (J. Li et al., 2015).
WP8 has also developed methodologies of assessing DNA sequence variants for their likelihood to be deleterious and increase disease risk. Web-based software has been developed for this purpose (M. J. Li et al., 2015), building on previous work (Li et al., 2013), using logistic regression to integrate information from multiple predictive software programs. Related to this and the previous section, is a review of association testing for multifactorial diseases and power for gene-based common variant tests and gene-based rare-variant burden tests (Sham and Purcell, 2014).
WP8 has developed and performed genome-wide complex trait analysis, and used this methodology to estimate the genetic and environmental components of variances and co-variances for multivariate phenotypes, something the available software available to date only does for the bivariate case.
Current genome wide association studies utilize SNP arrays that assay only a proportion of all SNPs in the genome, and rely on SNP imputation to capture untyped SNPs. As this approach will result in loss of statistical power in some circumstances, a full likelihood approach has been developed for association testing of an untyped marker by utilising an phased/unphased reference panel.
Association analysis of family data is complicated by the correlations among related individuals. WP8 is in the final stage of completing a methodology for using latent variable predictions from phenotypic information on all pedigree members in a regression on putative risk factors such as SNPs or environmental factors (manuscript in preparation).
The method development will be extended in order to provide further tools for testing environmental interaction with polygenic risk scores, and effect size distribution of interactions. Further implementation of methodologies for GxG and GxE will continue as soon as the new methods are fully developed.
While the development and implementation of the full range of statistical methods will continue after the project period, WP8 researchers have utilised the clinical and imaging datasets within EU-GEI to perform preliminary interaction analyses, e.g. on psychosis by age interaction for MRI data. Furthermore, GxE (including polygenic risk score analyses) and ExE analyses have been performed within EU-GEI. The tools and preliminary analyses generated by WP8 form the basis for full analyses on the entire EU-GEI dataset, which will be fully complete once the GWAS data on the entire EU-GEI dataset have been integrated with the environmental and clinical datasets.
WORK PACKAGE 9: ETHICS
Guidelines for harmonisation of procedures in schizophrenia studies in the EU
An important first step in any research study involving human subjects is confirming that the potential participants are capable of making an informed decision whether or not to enroll in the study, after being told of all procedures, risks and benefits. Known as providing informed consent, the process helps ensure that participants are acting of their own free will, that they understand what they are agreeing to do, and that they will be treated ethically during the study. But what happens when an investigation aims to involve people with an illness, such as schizophrenia, that can affect decision-making abilities? How can one be sure these individuals really have the capacity to decide for themselves?
WP9 has studied and monitored this ethical issue. On the basis of the study of the international and supranational dispositions in force in the field of research and experimentation with human beings, particularly the ones related to persons not able to consent (directed to the protection, among others, of persons with mental disorders) the following criteria in the field of research with patients with schizophrenia can be set:
1) The general rule for research with human beings is the need for informed consent by the research subject in application of the principle of autonomy.
2) A correct understanding of the mentioned bioethical principle of autonomy implies that not every human being is capable of self-determination. Some individuals lose this capacity wholly or in part because of mental disability as is sometimes the case in patients with schizophrenia. In any case, the extent of protection afforded to incompetent patients should depend upon the risk of harm and the likelihood of benefit, and the judgment that any individual lacks autonomy should be periodically re-evaluated and will vary in different situations.
3) In the case of patients with schizophrenia, the general rule will be to consider them competent. The different documents and conventions at the international level (Declaration of Helsinki, Belmont Report, Convention of Human Rights and Biomedicine of the Council of Europe and its Additional Protocol on Biomedical Research) do not rule out the possibility of undergoing research with incapacitated people (as long as this research is absolutely necessary so as to improve the condition of the persons suffering from these illnesses) and even accept expressly the possibility of research.
4) Notwithstanding the previously stated, in order to undergo research with persons not able to consent, some considerations have to be made:
a. The informed consent of the representatives or of a legal body will be required. In any case, the individual concerned shall as far as possible take part in the authorization procedure. No research can be carried out with an incompetent subject when the person objects to it.
b. Research on incompetent persons only can be done when the results of the research have the potential to produce real and direct benefit to his or her health or exceptionally and under the protective conditions prescribed by law, when the research has the aim of contributing, through significant improvement in the scientific understanding of the individual's condition, disease or disorder, to the ultimate attainment of results capable of conferring benefit to the person concerned or to other persons in the same age category or afflicted with the same disease or disorder or having the same condition and the research entails only minimal risk and minimal burden for the individual concerned.
5) From a principal point of view research with persons with mental disorders can be considered not only admissible but necessary as it promotes the improvement in the condition of those patients (principle of justice).
Gender in research
Schizophrenia is more common in men (Aleman et al., 2003), and research suggests that the prognosis is also worse in men (van Os & Allardyce, 2009). The reasons for these differences are not clear (Navarro et al., 1996), but the observation that neurodevelopmental disorders including autism, ADHD and dyslexia in general are more common in men suggests that the male developing brain may be more vulnerable to environmental insults (Castle & Murray, 1991) that, in interaction with predisposing genes, may result in a specific neurodevelopmental phenotype. EU-GEI researchers analysing data in the GROUP sample confirmed that gender is independently associated with an earlier onset of psychotic illness (Dekker et al., 2012b). In addition, male sex predicted longer duration of untreated psychosis predicting poorer prognosis (Apeldoorn et al., 2014), particularly in terms of quality of life (Boyette et al., 2014a). More analyses on gender differences are currently conducted, focussing on cognition, gene-environment interactions and outcome.
Gender in the EU-GEI consortium
The consortium has aimed at having an equal gender distribution. Women have been represented in the scientific management of the project, as work package leaders who serve in the Executive Board, as well as in all boards and committees of the project. Thus, at the senior level of the consortium scientific management (WP leader / vice leader) there have been five women (Myin-Germeys, Valmaggia, Emaldi, Velthorst and Parellada) representing 23% (5 out of 22) of senior scientific management. Although far from representing gender equality, we have considered this to be encouraging, given the still unbalanced gender ratio among senior schizophrenia researchers.
The overall gender ratio in EU-GEI, taking into account all researchers in the consortium, has been considerably higher at about 48% which has exceeded the average share of women in the academic career rank as experienced researchers in most of the member states according to the report of the Helsinki group on Women and Science about national policies on women and science in Europe (2002).
WORK PACKAGE 11 – TRAINING
An essential part of the activities of the training work package has been the development of an enclosed, and official web based training area of the EU-GEI website: GET-THERE (Gene Environment Tools - Training Home Education Reliability Europe).The website was intended to provide all EU-GEI researchers with information and training for the instruments to be used in the EU-GEI project for the diagnostic assessment, assessment of clinical characteristics and environmental risk factors. Apart from training documentation, manuals, instruments, score sheets and frequently updated questions and answers on procedures or inclusion criteria, it presented audiovisual material on the most advanced instruments assessed in Workpackage 2, 5, and 6.
In sum, GET-THERE included information on:
1. Manuals and guidelines for all instruments
Instruction videos (Family Interview of Genetic Studies, WAIS-III, Genetics, Premorbid Assessment Scale), Lectures (neuropsychology, SIS-R, CEQ, Combined Scales, OPCRIT, SDS, CAPE, Stigma-scales, Migration History, Experienced Sampling, CAARMS, In- and Exclusion criteria) for a comprehensive overview of selected instruments for WP2/WP6 and WP5 see below,
2. Eight lectures on Gene x Environment interactions in schizophrenia, presented by distinguished researchers in the field,
3. Background information on the instruments,
4. (English subtitled) training videos of the most advanced instruments (SIS-R, OPCRIT, CAARMS, GAF, List of Threatening Events, Childhood Experiences of Care and Abuse, Bullying),
5. Inter-Rater Reliability-measurement videos
6. Answers to Frequently Asked Questions,
7. Written practice exercises (NOS-DUP, GAF, Socio-Economic Status assessment procedure).
All training- and instruction videos have continuously been available on the website, and researchers have been kept up-to-date on difficulties around the assessment of instruments through the Frequently Asked Questions that have regularly been updated on the web.
Apart from (information on) the instruments, information on informed consent and inclusion procedures, was frequently updated. Through an email system including all researchers involved in the project, questions on recruitment procedures, and questions about whether someone was or was not eligible to participate in EU-GEI were frequently circulated. In this way, everyone was being kept involved and being kept up-to-date about the consensus rules of EU-GEI. Furthermore GET-THERE included DNA instructions on how to store, process, and extract material for DNA-processing. Also training in the use of PsyMate/ Experienced sampling protocol as part of WP5 was available online.
Assessment of rater reliability
One of the main targets of the training website has been to establish good inter-rater reliability. Working with numerous researchers from different countries and various disciplines is challenging. To generalize and compare findings across centers and countries, a key issue in this type of research has been the absolute need for uniformity in experimental procedures and experimenter behavior. To this means, an extensive, repetitious training procedure and reliability check was constructed. Training videos of the most advanced instruments were updated regularly.
For each of the training videos, a ‘golden standard score’ was determined through independent rating of the training videos by independent experienced researchers in Maastricht and Amsterdam. In case of disagreement, the head of the training work package was consulted.
Per instrument, we subsequently determined the maximum amount of errors/ deviation from the Golden Standard Score the researcher was allowed, in order ‘pass’ the video.
Researchers received feedback after rating the online videos. When researchers did not ‘pass’ a video, they were invited to try it again after a minimum of 24 hours. Rating of the online training videos was mandatory; only researchers that succeeded in passing the reliability checks were permitted to assess participants included in EU-GEI.
Rater reliability has been determined for: the CAARMS, List of Threatening Experiences, Bullying, SIS-R, OPCRIT and GAF-scale. A kappa of >.7 was considered acceptable.
Training made available for wider scientific community
A part of the EU-GEI training site has been made available to a broader public since the second project year. All training videos without confidential patient information are available for researchers who requested a password and user name via the EU-GEI website.
Demographic information researchers
At first entry to the GET-THERE website, every EU-GEI researcher was invited to fill out a short questionnaire on their age, place of birth, education, job title, etc. The demographic information of the researchers will be used for a further study regarding the influence of researcher characteristics on IRR ratings - in those who consented this information to be used for research (consent rate +/- 99%).
For complete text of S&T results, including publications, charts and tables, please see attachment:
“EU-GEI_final report_FINAL (29.06)”
The potential impact of the EU-GEI project is manifold.
1. The notion of gene-environment interplay as a researchable topic.
The study of gene-environment interplay has received an important as a result of EU-GEI activities. Gene-environment interplay at the level of EU-GEI was not undertaken before. It has brought about a close collaboration between traditionally genetics-focused researchers and researchers with a focus on environmental impact on risk. This collaboration will likely grow and intensify over the next decade as the data from EU-GEI will become fully available for hypothesis testing. A fully developed and validated environmental assessment battery is now available for the genetics research community. Analytical tools to tackle gene-environment analyses have been developed for use by the wider research community. In addition, multimodal and multi-method approaches, from risk prediction instruments to experience sampling to comparative incidence studies to family studies to neuroimaging and EEG studies, for the examination of GxE have been developed and implemented in EU-GEI which will guide other researchers wishing to engage in gene-environment interplay research.
2. Gene-environment interplay and clinical practice.
We have some early indications of clinically relevant findings. First, we have shown that experience sampling methodology, as developed in EU-GEI, can predict clinical transitions through the analysis of intensive time series of mental states and contextual factors. Second, we have shown that the psychotogenic effect of environmental risk factors such as cannabis use may be moderated by specific genetic variants. Third, we have shown the effect of a range of environmental factors on the onset and persistence of psychosis, particularly childhood adversity. Fourth, we have shown that mental health services in European countries may need to plan capacity in relation to country-specific variation in incidence related to underlying environmental factors.
3. Potential impact related to forthcoming analyses.
The bulk of the main EU-GEI data will be analysed in the coming year which we expect will yield some major findings in the area of gene-environment interplay in psychosis. Notably, absence of gene-environment interaction will also be a major finding that will significantly impact current theories on the origin of psychosis. Replication of specific findings, for example on the interaction between variation in FKBP1 and AKT1 and, respectively, childhood adversity and cannabis use will have major implications for current theories of psychosis and will spurn specific treatment research.
Main dissemination activities
Articles published/accepted for publication in peer-reviewed journals
During the five years of the project 95 articles have been published (of which 39 provided in Open Access). Peer-reviewed journals include: Schizophrenia Bulletin, Nature, Molecular Psychiatry, Human Genetics, British Journal of Psychiatry, PLoS One, Schizophrenia Research, BMC Psychiatry, Archives of General Psychiatry, Psychological Medicine.
In Year 5 the first EU-GEI core paper was published in Schizophrenia Bulletin “Identifying Gene-Environment Interactions in Schizophrenia: Contemporary Challenges for Integrated, Large-scale Investigations” (DOI: 10.1093/schbul/sbu069) using the established EU-GEI group authorname, including 187 researchers.
All published articles can obtained via the EU-GEI website. Publication of articles will continue for the upcoming years as more results become available. See also the list of planned synopses (per M60) in the annex.
Building relationships with external stakeholder groups
EU-GEI partners have continually included stakeholder groups by organising meetings and conferences for and with these groups. The meetings have included the following stakeholders and topics:
• Joint WP1-WP9-WP10 workshop: Social Inequality and Psychosis: changing perspectives, involving Spanish patients and relatives. Topics: destigmatising mental health, legal implications for schizophrenia research, experimentation and research with mentally ill people: legal aspects and fundamental rights
• Joint WP1-WP7 thematic workshops & events: “Psychose in Beeld”, involving Dutch and Belgian patients and relatives (Ypsilon), national health organisations, practitioners, pharmaceutical industry and general population. Topics covered were: Psychosis: science in practice, early intervention, new treatments for severe mental illness, discussion evening “Explaining Psychosis”.
• The Italian Research Experience (incl. WP2- P27.UNIPA) involving researchers and families of patients, Italian Family and Caregivers Group. Topics covered were: Update on drug development, understanding schizophrenia today, cognitive therapy, the sufferer’s perspective and a family forum.
• WP7/MUMC has been invited for workshops, master classes and other classes for medical specialists, psychologists and psychiatrists to present on Stigma (Resilience and Vulnerability, Recovery of symptoms and stigma, Stigma experiences). The workshops and classes were organised by the National Platform of Mental Health Care Services (LPGGZ) and regional mental health care services in the Netherlands
• WP2/UNIBO has organized training sessions covering Psychosis’ Risk Profile for Community Mental Health Centres (CMHCs) in Bologna. Training sessions specifically for mental health operators (psychiatrists, nurses, educators, psychologists and social workers) involved in the care of patients with first episode psychosis (FEP) to further explain the EUGEI study's aim and to share with the clinicians the “Psychosis’s Risk Profile”
• WP2/WP6/SERMAS has organized conferences with ALUSAMEN (Asociacion en lucha por la Salud Mental y los Cambios Sociales) and Fundacion Manantial to inform and train psychologists, psychiatrists and social workers on neuropsychological tests.
Furthermore a number of patient- and family related meetings have been attended by researchers and members of the External Advisory Board, including:
- European Network For Mental Health Service Evaluation- ENMESH
- EUFAMI Congress
- International Network Philosophy of Psychiatry
As part of scientific dissemination EU-GEI researchers have participated in international scientific conferences sharing their expertise in the areas of their respective Work Packages. At the first stage of the project the presentations have focused on the design of the study. As the project progressed first findings and preliminary results were disseminated.
International scientific conferences and meetings during the project have included:
- Congress of the European College of Neuropsychopharmacology
- World Congress of Psychiatry
- Congress of Biological Psychiatry
- International Symposia of Advances in Psychiatry
- European Congress of the European Psychiatric Association
- Schizophrenia International Research Conferences
- Congress of the European College of Neuropsychopharmacology
- International Congress in Schizophrenia Research
- World Congress of Psychiatric Genetics
- Conference of the Molecular Psychiatry Association
- International Congress of the Royal College of Psychiatrists
- Congress of the Psychiatric Association of Turkey
- European Congress of Psychiatry
To share the appraisal of current scientific knowledge on schizophrenia, including new findings emerging from EU-GEI, the project has issued several digital newsletters, covering scientific and other topics:
- Assessing environmental exposure: rural vs urban settings (Craig Morgan, WP2)
- TwinssCan study in Belgium (Jeroen Decoster and Ruud van Winkel, WP4)
- Assessing stress sensitivity in daily life - experience sampling research in EU-GEI
(Uli Reininghaus, Eva Velthorst, Barnaby Nelson – WP5)
- Looking at genetics from an environmental perspective (Alex Richards – WP3)
- City living and urban upbringing affect neural social stress processing in humans
(Andreas Meyer-Lindenberg WP4)
- Clinicopathological significance of psychotic experiences in non-psychotic young people
(Ian Kelleher and Mary Canon, WP5)
- Life Events and Psychosis: a Review and Meta-analysis (Stephanie Beards – WP2)
• Young researchers and their work in the project:
- Eva Velthorst (WP2 and WP11)
- Matthew Kempton (WP5)
Xplore Health – video
Xplore Health is a European educational portal on cutting-edge health research that offers innovative multimedia and hands-on experiences to young people through the internet, schools and science centres and museums. Xplore Health aims at bridging the gap between research and education, inspire future researchers and stimulate dialogues
In RP3 a team of Xplore Health prepared a 5-minute video, as part of a module on Mental Health, explaining state-of-the-art European research in Mental Health. The video explains what EU-GEI is, how researchers work, what results the project expects to get in the future and its significance for EU citizens.
EU-GEI has developed a unique set of instruments used for the diagnostic assessment, assessment of clinical characteristics and environmental risk factors. A web based training area of the EU-GEI website was developed: GET-THERE (Gene Environment Tools - Training Home Education Reliability Europe) to provide information and training for all EU-GEI researchers using the set of instruments.
A part of the EU-GEI training site has been made available since the second project year. All training videos without confidential patient information are available for a wider research-public
Overview EU-GEI training / public domain
INSTRUMENTS TRAINING TOOLS
CAARMS / Comprehensive Assessment of at Risk Mental States
instruction video, training video, reliability measurement, NEURAPRO vignette
Combined Social Scales
Sociodemographics Schedules, List of Threatening Experiences, Childhood Experiences of Care and Abuse, Bullying Questionnaire, Discrimination - Brief Impact of Events - Social Assessment tool
Inter rater reliability video
GAF / Global Assessment of Functioning
OPCRIT / Operational Criteria Checklist for Psychotic and Affective Illness
PAS / Premorbid Adjustment Scale
Cannabis Experience Questionnaire
Video of lecture
Shortened WAIS / Wechsler Adult Intelligence Test
SIS-R / Structured Interview for Schizotypy (Revised)
Presentation, questionnaires and background information
Momentary Assessment Technology
The concept of Momentary Assessment Technology has been patented already in 2010, with the support of BioMedbooster, a center of excellence for technology transfer, supporting the commercialization of new inventions and ideas (an initiative of MUMC and Industrial Bank LIOF).
In the meantime the (1st generation) hardware platform to apply the Momentary Assessment Technology has been developed by Wingz and is being employed in the EU-GEI project (WP5). The device has also been used in trials linked to WP4, WP5 and WP7. In year 4 and 5 of the project (SMEs Wingz and EIB and MUMC) have continued working on the development of a 2nd generation PsyMate: an App available for Android and iOS devices. This so-called PsyMate App is basically a platform to run ecological data-collection schemes and that can be customized to different needs in research and basic clinical practice.
Risk assessment charts
The other main deliverables from the project will be used for valorisation with non-commercial purposes, including the risk prediction charts:
- risk assessment chart for use in the prediction of onset schizophrenia spectrum disorders (WP2)
- risk assessment chart for transition from a prodomal state to a clinical psychotic disorder (WP5)
- risk assessment chart for variation in course and outcome in schizophrenia (WP7)
The risk prediction charts will become freely available via the EU-GEI website.
The project has produced two tools which are available via the EU-GEI website, providing guidelines and recommendations regarding:
- the organization of a user-led study on stigma resilience
- the organization of a web-based Training Program in multi-center, international psychosis research
The Software package PLINK is an open-source whole genome association analysis toolset, designed to perform a range of basic, large-scale analyses in a computationally efficient manner, developed, amongst others, by Dr Shaun Purcell and Prof. Pak Sham (P21 UHK). Prof. Pak Sham has developed, in EU-GEI, a module in PLINK to model GWAS x Environment and Common CNV x Environment interactions allowing us to use genome-wide approaches to test for environmental sensitivity (see WP8: GxE Data and Statistics).
Part of the software is already available via the web (http://hpcf.cgs.hku.hk:3838/drc/current ).
The developed methodology is set up to predict risk of Schizophrenia on the basis of
- an ACE multifactorial disease model, e.g. lifetime risk =1%, heritability=80%
- known risk factors, including polygenic score
- family history
List of Websites:
Address of project public website and relevant contact details
Scientific representative of the project's coordinator and organisation:
Prof. Jim van Os - UNIVERSITEIT MAASTRICHT
Vice-coordinator: Bart Rutten
tel: +31 43 3881263
Project website address: www.eu-gei.eu
Grant agreement ID: 241909
1 May 2010
30 April 2015
€ 15 030 922,89
€ 11 616 855
Deliverables not available
Grant agreement ID: 241909
1 May 2010
30 April 2015
€ 15 030 922,89
€ 11 616 855
Grant agreement ID: 241909
1 May 2010
30 April 2015
€ 15 030 922,89
€ 11 616 855