Lisa Dinkler

Avoidant/restrictive food intake disorder (ARFID) is an eating disorder characterized by severe food restriction unrelated to weight or shape concerns. Affecting 1–2% of the population, it typically emerges in early childhood, is often chronic, and leads to poor health and high healthcare costs. Despite its impact, ARFID remains underdiagnosed and undertreated owing to its heterogeneity, unclear diagnostic boundaries, and lack of structured care pathways. This project aims to improve ARFID’s conceptualization and identification to enable timely, targeted interventions and prevent chronicity.I will leverage unparalleled genomic, survey, and register data from ~10,000 individuals with and ~32,000 without ARFID across six countries. Using epidemiological and novel, genetically informed classification methods, I will (1) identify risk factors and develop a risk prediction index to improve detection and distinguish ARFID from transient childhood eating difficulties, (2) deconstruct ARFID heterogeneity by identifying clinically meaningful subtypes and refining its diagnostic boundaries, and (3) translate findings into a structured clinical decision tool to support clinicians in diagnosis and patient triage.By refining ARFID classification, enhancing clinical decision-making, and informing targeted interventions, this work will improve access to care and patient outcomes. A key strength is the extensive national and global collaboration supporting this effort.

This project targets tertiary prevention in psychiatry: identifying, early in the course of care, individuals most likely to develop severe, persistent illness and tailoring care accordingly. We focus on major depressive disorder (MDD), anorexia nervosa (AN), and bulimia nervosa (BN)—disorders with high female predominance and major contributions to disability among Swedish women. Using Sweden’s exceptional national health registers, we will construct large, richly phenotyped cohorts spanning four decades and apply modern machine learning to detect reproducible, data-driven patterns of long-term outcomes (Aim 1), identify early predictors of poor trajectories including clinical, social, developmental, and genomic factors (Aim 2), and use causal inference to determine which treatments are associated with better outcomes in those at highest risk (Aim 3). Analyses will emphasize replication, generalizability, and clinical relevance, with a focus on sex differences. Preliminary results demonstrate feasibility and identify clinically actionable subgroups, including those at elevated risk for suicide or diagnostic crossover. Successful completion will yield robust, generalizable predictive models to guide personalized, proactive psychiatric care and inform scalable tertiary prevention strategies. Findings will be directly relevant to Swedish healthcare and may serve as a model internationally.

Between 2014 and 2024, a multidisciplinary team of investigators at the Centre for Eating Disorders Innovation (CEDI) at Karolinska Institutet collected extensive clinical and self-report data and biological samples for genomic and other -omic studies from over 24,000 individuals with eating disorders (anorexia nervosa, bulimia nervosa, binge-eating disorder, atypical anorexia nervosa, and avoidant restrictive food intake disorder) and appropriately matched controls. CEDI studies were funded by Swedish (e.g., Swedish Research Council, FORTE, Hjärnfonden, ALF)

Danish (Lundbeckfonden)

American (National Institute of Mental Health)

industry (Shire/Takeda)

and philanthropic sources (e.g., Klarman Family Foundation, Actum Foundation)—an unprecedented investment in eating disorders science.CEDI data and samples represent a global resource that is poised to answer future questions about the biology, psychology, and environmental factors that influence risk, maintenance, and recovery from eating disorders and co-occurring conditions such as depression, anxiety, and obsessive-compulsive disorder. Moreover, our discovery of a metabolic genetic component to anorexia nervosa lends our data and samples to investigations of conditions of global public health relevance such as obesity and type 2 diabetes. No eating disorders collection in the world rivals the depth and completeness of our resource. To ensure its longevity, utility, and accessibility in accord with FAIR principles, we propose the development of CEDIX—a comprehensive, harmonized, searchable database comprising data and metadata from all CEDI studies linked with a searchable and accessible biobank of DNA, saliva, microbiome, plasma, and other study-specific biological samples housed in the KI Biobank. We will develop and systematize a self-supporting scheme for requesting data and sample access

for verifying ethical approval and GDPR compliance

and a mechanism for growing CEDIX by requiring deposit of created variables and novel findings back into the database. FORTE support would provide the personnel and infrastructure to refine, thoroughly test, and publicize availability of the resource. CEDIX and its biorepository will be an invaluable resource for current researchers as well as future global investigators who develop novel hypotheses and apply emerging methodology to address etiology, treatment, recovery, and relapse of these life-impairing and too frequently fatal conditions.

We propose the Eating Disorders Genetics Initiative 2 (EDGI2), a new collaborative R01 in response to PAR-23-050 Clinical Studies of Mental Illness. Its predecessors, EDGI1 (R01 MH120170) and ARFID-GEN (R56 MH129437), were highly successful. We unite the three global EDGI1 sites to advance genomic discovery across major eating disorders (EDs) and identify biologically, clinically, and therapeutically actionable insights. Inclusion of foreign sites is essential to achieve the large sample sizes required for genetic discovery

these discoveries will improve understanding and treatment of EDs in the U.S. Aim 1: EDGI2 will increase sample size, geographic origins, and ED phenotypes. Using our harmonized online assessment battery, we will phenotype and bio-sample 20,000 new participants with anorexia nervosa (AN), bulimia nervosa, binge-eating disorder, avoidant/restrictive food intake disorder, and controls, over-sampling severe and enduring AN, whose DNA may be enriched for causative alleles. Aim 2: We will explicate ED heterogeneity and biology by conducting standard genomic analyses (GWAS of diagnoses, trans-diagnostic behaviors, and continuous traits, PRS, and rare variant CNV)

identifying clinically meaningful subgroups

and evaluating that AN is a metabo-psychiatric disorder using LDSC, PRSet, pheWAS, and Mendelian randomization to clarify causal direction. Aim 3: We will assess genetic and environmental risk and resilience factors to inform prediction by phenotypically characterizing cases and controls with high and low PRS for EDs and then by genotypically characterizing those with severe phenotypes. We will identify genetic or molecular patterns across cases and controls and phenotypically characterize identified subtypes. Aim 4: To determine where in the body EDs “live”, we will identify brain cell types and anatomical regions implicated by genomic studies of EDs

predict genetically regulated gene expression (GREx) in brain, gut, adipose, and other ED-relevant tissues

use snRNAseq atlases to sharpen preliminary GTEx and TWAS analyses to identify brain cell types strongly implicated by the genomics of each ED

expand to relevant non-brain cell types (e.g., adipose, muscle, liver, salivary)

and use dynamic GREx to model gene expression in ED-relevant contexts (e.g., sex, BMI, stress), enabling precise and personalized modelling of gene expression. Aim 5: EDGI2 will culminate in a Translational Summit uniting forward-thinking stakeholders from multiple sectors to develop a translational roadmap for evidence-based ED prevention and treatment. EDGI2 will yield critical knowledge about genetic and environmental risk for EDs, reveal mechanisms that potentiate or protect against genetic risk, and transition ED genetics from discovery to clinical translation.

Research problem and specific questions. Avoidant/restrictive food intake disorder is a serious, taxing, and potentially life-threatening condition that affects 1-5% of the population. First recognized in 2013, we know little about causes, course, maintenance factors, and outcome

however, nutritional, medical, and psychosocial consequences are severe. We propose establishing a re-contactable, longitudinal cohort (ARFID InitiativE Sweden: ARIES) to study genetic and environmental contributors to ARFID through a developmental lens. ARIES will include rich phenotyping, DNA samples, and stool samples in a national data- and biorepository linkable to Swedish population health and quality registers to answer fundamental questions about this pernicious, life-impairing, and understudied condition. Data and method. We will first form a community-based Parent Advisory Committee (PAC) to guide our assessment domains and then recruit 500 children ages 4-9 with ARFID and 500 age and gender matched controls

collect parental report via online questionnaires on development, feeding history, food preferences, behavior, mood, feeding-related adverse events, and other domains suggested by the PAC

and create a biobank of saliva (for DNA) and stool (for intestinal microbiota and microbiome). Plan for project realisation. We will launch a national recruitment campaign using our country-wide collaborations with eating disorders clinicians and pediatricians and our social media channels to reach parents of children with ARFID. After an online eligibility screen, parents will complete questionnaires

submit saliva samples for their children

and provide a stool sample. Families will enroll in ARIES, agreeing to regular follow-ups. In addition to baseline research aims, ARIES will enable identification of predictors of transient versus enduring illness course, impact of puberty on ARFID symptoms, and impact of ARFID on educational achievement. Linking ARIES with population registers will allow research questions about ARFID comorbidity and healthcare utilization. Relevance. No evidence-based treatment for ARFID and no national guidelines exist. Parents desperately seek guidance on managing ARFID and fear for their child’s long-term well-being. ARIES will be the foundation for understanding ARFID, identifying factors related to onset, maintenance, and outcome, and be a national and international resource for essential research.