Major genetic study links cannabis use with psychiatric, cognitive, and physical health

The latest analysis by researchers from the UC San Diego School of Medicine in collaboration with the company 23andMe has identified precise genomic regions associated with the likelihood of people trying cannabis and the frequency of their subsequent use. This is one of the largest studies to date combining genotype data and self-reported behavior, which further dissected the relationships with psychiatric, cognitive, and somatic health traits. The findings suggest that cannabis-related behaviors are polygenic in nature and share some genetic architecture with mood, anxiety, and psychotic disorders, as well as with measures of executive function, impulsivity, and risky decision-making. For public health, this opens up the possibility of earlier recognition of risky patterns, as well as thoughtful targeting of interventions for people who are developing habits of more frequent use in the early stages.

Why the timing is important and what it means for public health

The last decade has seen an increase in the frequency of cannabis use and a proliferation of products that vary in potency, method of administration, and cannabinoid combinations. With rising consumption, public health systems are striving to better identify risk and vulnerability factors. While much has been written about environmental and behavioral influences, the genetic contribution to risk has long been less clear to the general public. This study delivers an important message: genetic differences do not determine destiny, but they can explain part of the variability in who will try cannabis, how quickly the frequency will increase, and who will develop patterns that interfere with daily functioning. In combination with education, psychosocial measures, and market regulation, these findings offer a tool for more precise, evidence-based preventive programs. It is particularly relevant to emphasize that risk should not be judged by genetics alone: context, age of onset, environment, and mental health collectively form an "ecosystem" of risk.

Design and methodology: GWAS on "ever used" and "frequency"

The research team conducted a genome-wide association study (GWAS) on two related but conceptually distinct variables. The first is the binary measure "ever used cannabis" (lifetime use), which captures the early, initial contact. The second is "frequency of use," a quasi-quantitative measure that better describes continuous behavior and habits. Participants answered via an online questionnaire whether they had ever consumed cannabis, and those who had, further estimated how often. Genotypic data were analyzed with strict controls for population stratification, data quality, and potential confounders. Statistical models accounted for standard covariates, after which the obtained associations were verified and extended through secondary analyses on independent datasets, including large biobanks. In addition to the primary GWAS, polygenic risk scores (PRS) were created and phenome-wide association studies (PheWAS) were conducted to track links with thousands of clinical and behavioral traits.

Key points: CADM2 and GRM3 at the center of the network

The results confirmed two genes that stand out as central nodes of the biological story. Cell Adhesion Molecule 2 (CADM2) is associated with both whether someone will ever try cannabis and how often they will use it. CADM2 is involved in establishing synaptic connections and fine-tuning communication between neurons, and in previous works, it has been consistently linked to impulsivity, risk-taking, and body mass index. Metabotropic glutamate receptor 3 (GRM3), on the other hand, is part of the glutamatergic system, crucial for synaptic plasticity and learning. This gene has been linked in multiple studies to psychoses, particularly schizophrenia and mood disorders, and here it shows a connection with cannabis-related behaviors. In other words, at the heart of the findings are pathways that regulate communication between neurons and systems that shape executive functions and decision-making.

The expanded genetic landscape: new names and an alliance of old acquaintances

In addition to CADM2 and GRM3, a number of additional genes and regions have been identified that together build the polygenic architecture of these behaviors. For lifetime cannabis use, several dozen genes have been highlighted, while for frequency, more key loci have been singled out, among which the area around CADM2 re-emerges. A significant portion of these genes has not been explicitly linked to cannabis until now, which reinforces the impression that we are mapping new territory and opening topics for functional studies. What connects them is their involvement in neurodevelopmental processes, synaptic shaping, and modulation of neurotransmitter systems, and the common pattern also includes a tendency towards impulsivity, a greater demand for reward, and sensitivity to environmental stimuli.

How much is "genetic"? Heritability estimates and polygenic risk

Both outcomes examined show a measurable, though moderate, component of heritability at the whole-genome level. This means that both "ever use" and "frequency of use" are polygenic – traits determined by many variants of very small individual effect. At the individual level, polygenic risk scores (PRS), calculated from these GWAS results, explain only a small part of the variance, but at the population level, PRS become useful tools for observing patterns and testing hypotheses. When PRS is linked to thousands of phenotypes in independent clinical cohorts, we get a "topographical map" of risk that indicates which psychiatric, cognitive, and somatic traits most often overlap with a genetic predisposition to cannabis use.

Genomic correlations with more than a hundred traits: psychiatry, cognition, and somatics

Genome-wide correlation analyses have shown that the genetic predisposition to trying cannabis and to more frequent use overlaps with a range of psychiatric and cognitive measures, as well as with several important physical health traits. In the domain of mental health, schizophrenia, ADHD, anxiety, depression, and bipolar disorder stand out. In the cognitive field, links are visible with executive functions, information processing speed, risk-taking, and impulsivity. In the area of somatic conditions, connections are found with diabetes, chronic pain, and coronary heart disease. Such a pattern suggests that early contact and more frequent use do not occur in isolation but are part of broader behavioral and biological circuits that shape how a person makes decisions, regulates emotions, and responds to stress.

Links with risk behaviors: tobacco, infections, and autoimmune diseases

Polygenic scores for cannabis use also show links with smoking habits and other forms of nicotine consumption, which is not unexpected given the shared dopamine and control circuits in the brain. In hospital biobanks, a signal for an increased risk of certain infectious diseases, including HIV and viral hepatitis, was also observed. These are most likely indirect effects, through a network of behaviors, social determinants, and healthcare, rather than a direct biological effect of individual variants. Additionally, links with autoimmune diseases were discovered, which raises new questions about the role of immune pathways in drug use-related behaviors, but also about possible measurement and diagnostic biases that need to be carefully controlled in future replications.

From "first steps" to disorder: what behaviors before CanUD reveal

Most people who try cannabis will not develop a cannabis use disorder (CanUD). However, earlier onset and more frequent use are consistently associated with a higher risk of problems. That is why the focus of this study on behaviors that precede a clinical diagnosis is extremely important: early patterns may be the most suitable for interventions. If risky combinations are observed – for example, a higher polygenic risk, more pronounced impulsivity, and a rapid transition from experimentation to frequent use – it is possible to direct resources to counseling, monitoring, and support before a complex of symptoms develops that significantly disrupts school, work, or family relationships.

Methodological nuances: correlation versus causation and next-generation tools

LD-score regression is used to estimate the overlap of the genetic architecture of two traits, but it does not tell us about the direction of effect. The question of causality requires additional tools. Mendelian randomization (MR), which uses polygenic instruments, can help test the potential effect of one trait on another, while longitudinal cohorts with detailed phenotyping can disentangle the sequence of events in real life. In the next phases, a combination of MR with measures from digital biomarkers (e.g., passive monitoring of behavior via smart devices), integration with transcriptomics, and mapping the expression of candidate genes in relevant brain tissues is expected. Such approaches will allow for a more precise description of the mechanisms by which variants in CADM2, GRM3, and related genes influence behavioral trajectories.

Glutamatergic gates: why GRM3 is a potential therapeutic target

Glutamate is the main excitatory neurotransmitter, and metabotropic receptors – including the one encoded by GRM3 – fine-tune synaptic plasticity and learning processes. If early contact with cannabis is genetically linked to variants in GRM3, it raises the question of whether modulators of the glutamatergic system could help redirect habits or alleviate comorbidities that accompany more frequent use. Although it is too early to talk about drugs, knowing the involved pathways allows for faster and targeted generation of hypotheses for preclinical trials and for the "repurposing" of existing pharmacological mechanisms.

CADM2 and the "impulsive brain": links to everyday decisions

In recent years, CADM2 has emerged as a gene associated with decision-making, reaction speed, and preference for short-term rewards. This does not mean that variants in CADM2 "create addiction," but it indicates that certain neurobiological patterns may direct people towards more frequent testing of environmental stimuli, including psychoactive substances. If we combine this with an early onset, high-potency products, and a social context that supports use, we get a scenario where there is a higher risk for developing problematic patterns. Therefore, self-regulation interventions, reward delay training, and strengthening executive functions are a logical addition to educational programs.

Clinical practice today: what we have and what is missing

There are currently no approved pharmacotherapies for cannabis use disorder. The best results are achieved with psychosocial approaches: motivational interviewing, cognitive-behavioral therapy, contingency management, and integrated care for comorbidities such as depression, anxiety, or sleep disorders. For now, genetic information is an additional layer that can inform the conversation about risk, and in the future, it could help in triaging and personalizing interventions. For example, a person with a higher polygenic risk and early patterns of frequent use could benefit from more intensive monitoring and targeted support that includes techniques for impulse regulation and behavior planning.

Ethics and privacy: who is in the sample and how we protect data

Biobanks and consumer genetic platforms provide samples of the size needed to detect variants of small effect, but at the same time, they carry a risk of bias. Participants often come from specific demographic groups, so the results must be carefully generalized to populations that are traditionally underrepresented in research. In addition, data confidentiality and security remain a priority: transparent informed consent, oversight by independent ethics committees, and limiting the secondary use of data are key elements of a sustainable research framework. Only when a balance is maintained between scientific benefit and privacy protection will these findings have their full meaning in public health.

Croatian perspective and education system

In Croatia, cannabis is prohibited outside of strictly defined medical indications. However, experiences from school surveys show that experimentation exists, while frequent use is concentrated in relatively small groups. Genetic findings like these can help in designing content that combines cognitive training (self-regulation, planning, reward delay) with clear information about risks. School and local prevention programs need to be precisely targeted at vulnerable groups and take into account mental health, family dynamics, and social conditions that modify risk.

Research horizon: replications, diverse ancestries, and integration with neuroscience

The next chapters of this story require replications in samples of different ancestries, including multi-ancestry populations and younger age groups. Detailed functional studies are also needed to show how variants in CADM2 and GRM3 change neural circuits – from the level of expression to network dynamics. Integrated models that combine genetics, epigenetics, transcriptomics, and brain imaging could describe the "mechanistic bridges" between alleles and behavior. In parallel, it is important to standardize measures of cannabis use (dose, frequency, route of administration) and monitor the product profile so that future analyses can be accurately compared.

What we know today, October 15, 2025

The study published on October 13, 2025, confirms that CADM2 and GRM3 are reliable "anchor" genes for understanding early cannabis use. In addition to them, a number of additional loci have been identified, and polygenic scores show useful, although limited, predictive properties in populations. Genomic correlations with psychiatric, cognitive, and somatic outcomes strengthen the idea that these are complex, partly shared biological pathways. The scientific community now has a clearer blueprint of where to look for therapeutic targets, how to construct preventive programs, and how to connect genetics with behavior and clinical outcomes in real-world settings.

Glossary and quick reference

• GWAS — a genome-wide association study of variants across the genome with a particular trait in large samples.


• PheWAS — an examination of how the polygenic risk for one trait correlates with a multitude of other clinical and behavioral phenotypes.


• PRS — polygenic risk score; an aggregate indicator of genetic predisposition.


• CanUD — cannabis use disorder; a spectrum of problems that impair daily functioning.


• CADM2 — a cell adhesion gene associated with impulsivity and decision-making.


• GRM3 — a gene for a metabotropic glutamate receptor, involved in synaptic modulation and cognitive processes.

Who benefits most from these findings

Policymakers evaluating regulation and public health measures, clinicians who work with comorbidities daily, and teachers and parents who want to recognize risky patterns. For researchers, these data provide guidance on where to look for the next mechanisms and how to design studies that will bring genetics, behavior, and clinical outcomes closer together.