Browse the corpus

The human brain undergoes extensive remodeling during adolescence, with prefrontal cortex maturation continuing into the mid-twenties (, ). This extended developmental period involves synaptic pruning, white matter myelination, and circuit refinement processes that create both opportunities for learning and windows of vulnerability to environmental influences (, ). The endocannabinoid system plays critical roles in adolescent neurodevelopment, regulating neuronal migration, axon guidance, and synaptic pruning (). Cannabinoid CB1 receptors show peak expression in adolescent prefrontal and limbic regions during active developmental periods, creating potential windows of heightened sensitivity to cannabis exposure ().

Cannabis use among adolescents has increased significantly over the past two decades, with approximately 35% of U.S. high school seniors reporting past-year use in recent national surveys, and the average age of initiation decreasing to 14.6 years compared to estimates of 17–18 years in earlier cohort studies (–). Concurrent increases in cannabis potency, with contemporary products containing 15-25% THC compared to 3-4% in previous decades, may amplify risks for developing psychosis and other adverse neuropsychiatric outcomes (). Changing legal frameworks and decreased risk perception among adolescents have contributed to this trend, creating an urgent need to understand the neurodevelopmental consequences of adolescent cannabis exposure (–). Emerging evidence suggests that both timing and frequency of cannabis exposure during adolescence are critical determinants of neurodevelopmental risk (), with frequency of use potentially mediating the relationship between early initiation and adverse outcomes (). The “critical period hypothesis” proposes that cannabis use during adolescence produces qualitatively different and more severe effects compared to adult-onset use (–). This is supported by adolescent-specific brain vulnerability, distinct CB1 receptor developmental patterns, and epidemiological evidence showing worse outcomes with earlier versus later onset (, , ). Studies have documented structural brain alterations, cognitive deficits, and increased addiction risk specifically associated with adolescent-onset cannabis use (), with some effects persisting years after cessation (, ). However, findings are not entirely consistent, with some well-controlled studies questioning causal relationships and highlighting the role of confounding factors (, ).

Despite growing research interest, significant gaps remain in our understanding of adolescent cannabis effects. Many studies are cross-sectional or have limited follow-up periods, and heterogeneity in methodologies complicates synthesis. The observational nature of most research limits causal inference, as early cannabis users often differ from non-users across multiple domains, including pre-existing psychiatric vulnerabilities, adverse childhood experiences, family history of substance use disorders, socioeconomic disadvantage, and co-occurring use of other substances. These knowledge gaps have important implications for clinical practice and policy. Clinicians need evidence-based guidance for adolescent risk assessment and treatment, while policy debates require clear understanding of developmental vulnerability to inform age-based restrictions and prevention approaches. Given increasing adolescent cannabis use, changing legal frameworks, and emerging evidence for developmental vulnerability, a comprehensive synthesis of current knowledge is urgently needed. While previous reviews have examined aspects of adolescent cannabis effects, no recent systematic review has comprehensively evaluated evidence for developmental timing effects across multiple domains.

This systematic review aims to synthesize current evidence on the neurodevelopmental consequences of adolescent cannabis use, with particular focus on how effects differ by age of initiation. Our specific objectives are to: Examine neuroimaging evidence of structural and functional brain changes associated with adolescent cannabis use; Evaluate cognitive consequences and recovery patterns following adolescent versus adult-onset cannabis use; Assess addiction vulnerability and progression patterns in adolescent-onset users; Analyze long-term functional outcomes including educational, occupational, and mental health trajectories; Identify biological mechanisms underlying developmental vulnerability and individual differences; Evaluate methodological quality and identify limitations in current evidence; Provide evidence-based recommendations for clinical practice, prevention, and future research priorities. By comprehensively reviewing evidence across these domains, this systematic review aims to inform evidence-based approaches to adolescent cannabis prevention and intervention while identifying critical areas for future investigation.

This systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines ()(; ). The protocol for this systematic review was registered with PROSPERO (International Prospective Register of Systematic Reviews) (CRD420251165329) A comprehensive literature search was conducted using PubMed, EMBASE, PsycINFO, and Cochrane databases from January 2000 to September 2025. The search strategy employed a combination of Medical Subject Headings (MeSH) terms and free-text keywords related to: (1) cannabis exposure (“cannabis,” “marijuana,” “THC,” “cannabinoid,” “hashish”); (2) adolescent populations (“adolescent,” “teenager,” “youth,” “juvenile,” “young adult,” “developmental”); (3) neurodevelopmental outcomes (“neurodevelopment,” “brain development,” “neuroimaging,” “MRI,” “cognitive function,” “executive function,” “memory,” “intelligence”); and (4) timing factors (“early onset,” “age of initiation,” “adolescent-onset,” “developmental timing”).

Search terms were combined using Boolean operators (AND/OR) and adapted for each database’s specific syntax. Additionally, we conducted manual searches of reference lists from included studies and relevant review articles to identify additional eligible studies. Grey literature was searched through conference proceedings from major neuroscience and addiction conferences (Society for Neuroscience, Research Society on Alcoholism, International Cannabis Research Society) and doctoral dissertation databases (ProQuest Dissertations & Theses Global) ().

Population: Studies examining adolescent cannabis users (age of first use ≤18 years) or comparing adolescent-onset versus adult-onset cannabis users. Exposure: Clear definition of cannabis use patterns, frequency, duration, or age of initiation. Outcomes: Included at least one of the following measures: -Structural or functional neuroimaging (MRI, fMRI, DTI, PET) -Cognitive assessment using standardized neuropsychological tests -Longitudinal behavioral or functional outcomes (academic, occupational, mental health) -Biological measures related to neurodevelopment (receptor expression, neurotransmitter function) Study design: Cross-sectional comparisons, longitudinal cohort studies, or intervention studies. Sample size: Studies with adequate statistical power to detect clinically meaningful differences. For comparative studies, minimum 15 participants per group; for longitudinal studies, minimum 30 participants at baseline. Studies with smaller samples were considered if they presented particularly innovative methodologies or unique assessments not available in larger studies. All included studies met the minimum sample size criteria. For studies with total samples near the threshold (e.g., n=33), group distributions were verified to ensure adequate power in each comparison group (≥15 per group for case-control or cross-sectional designs).

Confounding control: Studies that controlled for or assessed key confounding variables including alcohol use, other substance use, and socioeconomic factors, or that provided sufficient data to evaluate the influence of these factors. Publication: Peer-reviewed articles published in English-language journals, plus high-quality grey literature (doctoral dissertations, government reports, conference abstracts) meeting rigorous methodological criteria. Methodological quality: Studies with clearly defined methodology and appropriate statistical analyses. Pilot studies or those with methodological limitations were included if they provided unique contributions to understanding adolescent cannabis use effects, with appropriate discussion of limitations in results interpretation. Animal studies were included exclusively when they provided direct mechanistic evidence pertaining to endocannabinoid system development during adolescence that could not be derived from available human studies. Given the focus of this review on human neurodevelopmental outcomes, preclinical studies were considered only as supplementary biological context and were therefore limited to studies most directly informing the critical period hypothesis examined her.

-Focused exclusively on acute intoxication effects or withdrawal symptoms without examining chronic use effects -Included participants with active psychotic disorders or severe psychiatric conditions at baseline that could confound neurodevelopmental measures -Lacked clear operational definitions of cannabis use patterns or age of initiation - Had follow-up periods less than 6 months for observational longitudinal cohort studies (this criterion did not apply to randomized controlled trials, where shorter intervention periods were acceptable given the experimental design and controlled conditions) -Were case reports, editorials, reviews, or studies without original data -Used exclusively preclinical animal models without human validation (except for mechanistic studies providing biological context) -Failed to adequately control for or assess major confounding variables -Had significant methodological limitations that compromised data quality or interpretation.

Two independent reviewers (initials) conducted the initial screening of titles and abstracts using predefined eligibility criteria. Full-text articles of potentially eligible studies were then independently reviewed by both reviewers. Disagreements were resolved through discussion, and when necessary, a third reviewer was consulted. Data extraction was performed systematically using a standardized data extraction form that captured: -Study characteristics: First author, publication year, study design, country, sample size, demographic characteristics -Cannabis use variables: Age of initiation, frequency of use, duration of use, potency measures (when available), abstinence periods -Comparison groups: Control group characteristics, matching criteria, group definitions -Outcomes: Neuroimaging findings (structural volumes, connectivity measures, activation patterns), cognitive test results, functional outcomes, biological measures -Methodology: Neuroimaging protocols, cognitive assessment batteries, statistical approaches, confounding variable control -Results: Effect sizes, confidence intervals, p-values, direction of associations ().

Study quality was assessed independently by two reviewers using the Newcastle-Ottawa Scale (NOS) ()adapted for cross-sectional and cohort studies (). The assessment evaluated three domains: -Selection of study groups: Representativeness of exposed cohort, selection of non-exposed cohort, ascertainment of exposure -Comparability of groups: Comparability on the basis of design or analysis controlled for confounders -Assessment of outcomes: Assessment of outcome, adequacy of follow-up (for cohort studies) Studies were rated as good quality (≥7 stars), fair quality (4–6 stars), or poor quality (≤3 stars). For intervention studies, the Cochrane Risk of Bias tool was used to assess randomization, allocation concealment, blinding, incomplete outcome data, selective reporting, and other potential sources of bias. Additional sources of bias specifically assessed included: -Selection bias: Differences between cannabis users and controls in baseline characteristics -Information bias: Recall bias in self-reported cannabis use, measurement error in neuroimaging or cognitive testing -Confounding bias: Inadequate control for alcohol, tobacco, other substances, socioeconomic status, genetic factors -Publication bias: Assessment through funnel plot examination and Egger’s test when sufficient studies were available

Given the heterogeneity in study designs, cannabis use definitions, and outcome measures, a narrative synthesis approach was employed rather than formal meta-analysis. Studies were grouped thematically by: Neuroimaging findings (structural and functional); Cognitive and academic outcomes; Long-term functional outcomes; Biological mechanisms and individual differences. For studies reporting similar outcomes with comparable methodologies, effect sizes were calculated and compared qualitatively. When possible, standardized mean differences were computed to facilitate comparison across studies. Subgroup analyses were conducted based on: -Age of cannabis initiation (early adolescence ≤16 years vs. late adolescence 17–18 years vs. adult onset >18 years)

-Frequency and intensity of use (occasional vs. regular vs. heavy use). Frequency of use categories were operationalized based on definitions provided by individual studies, which varied across the literature. For the purposes of subgroup analyses, we adopted the following working definitions: occasional use (fewer than once per week or ≤4 days/month), regular use (at least weekly but less than daily, or 5–19 days/month), and heavy use (daily or near-daily use, ≥20 days/month). When studies did not explicitly adopt these thresholds, categorization was based on the closest available definition reported by the authors. This variability in operationalization across studies is acknowledged as a potential source of heterogeneity and is discussed in the limitations section.

The methodological quality of included studies was rigorously evaluated using validated tools appropriate to each study design: Observational studies: The Newcastle-Ottawa Scale (NOS) was used to assess cohort and cross-sectional studies across three domains: selection of study groups (0–4 stars), comparability of groups (0–2 stars), and ascertainment of exposure/outcome (0–3 stars). Additional criteria specific to substance use research included: analytical confirmation of cannabis exposure (weighted higher than self-report), use of validated neuropsychological assessment tools, and adequacy of abstinence verification methods. Randomized controlled trials: The revised Cochrane Risk of Bias Tool (RoB 2.0) () was used to evaluate RCTs across five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of outcomes, and selective reporting. Each domain was rated as “low risk,” “some concerns,” or “high risk” of bias.

Grey literature: The AACODS checklist () was applied to assess grey literature across six domains: authority (credibility of authors and institutions), accuracy (methodological rigor and data presentation), coverage (scope and completeness of information), objectivity (presence of bias or conflicts of interest), date (timeliness and relevance), and significance (relevance to the research question). Each domain was scored from 0–2 points (maximum score 12). Grey literature sources scoring ≥8 points were considered high quality, 6–7 points moderate quality, and <6 points low quality. Only moderate-to-high quality grey literature was included in the final synthesis. Quality classification: Studies were classified as high quality (7–9 NOS stars or low risk of bias across all RoB 2.0 domains), moderate quality (4–6 NOS stars or some concerns in 1–2 RoB 2.0 domains), or low quality (0–3 NOS stars or high risk of bias in multiple RoB 2.0 domains). Two independent reviewers conducted the quality assessment, with disagreements resolved through discussion and consultation with a third reviewer when necessary. Particular attention was paid to biochemical verification of abstinence, use of validated cognitive assessment instruments, and control for potential confounding factors including polysubstance use and psychiatric comorbidity.

Sensitivity analyses were conducted to assess the robustness of findings across different study characteristics: Examining results when including only high-quality studies (Newcastle-Ottawa Scale ≥7 stars) to determine whether findings remain consistent when restricted to the most methodologically rigorous studies. Comparing findings from longitudinal versus cross-sectional studies to assess whether study design influences observed associations. Examining studies with different age cutoffs for “early onset” (≤16 years vs ≤17 years vs ≤18 years) to evaluate consistency across operational definitions. Comparing results from studies with biochemical verification of cannabis use versus those relying solely on self-report to assess the impact of measurement validity. Given the heterogeneity in outcome measures precluding formal meta-analysis, publication bias was assessed through: (1) systematic examination of grey literature sources including doctoral dissertations and conference proceedings; (2) qualitative assessment of study characteristics that might influence publication likelihood; and (3) examination of study registration and protocol availability where possible.

The systematic search identified 3,421 potentially relevant records across multiple databases. Following PRISMA 2020 guidelines, after removing 1,247 duplicates and applying inclusion/exclusion criteria through title and abstract screening, 156 full-text articles were assessed for eligibility. Subsequently, 120 studies were excluded for the following reasons: lack of adolescent-specific focus or age-of-onset data (n=42), insufficient sample size (n=28), lack of neuroimaging/cognitive outcomes (n=21), exclusively acute effects studies (n=15), inadequate confounding control (n=8), active psychotic disorders at baseline (n=5), and failure to meet inclusion criteria upon re-evaluation (n=1). Thirty-six studies meeting all methodological requirements were included in the final systematic review ().

Study Design Distribution: The included studies comprised various methodological approaches: longitudinal cohort studies with repeated assessments (n=23, 62.2%), cross-sectional comparisons examining cannabis users versus controls (n=8, 21.6%), randomized controlled intervention studies (n=4, 10.8%), and case-control studies with retrospective analysis (n=2, 5.4%). For studies where original authors did not report quantitative effect sizes or precise numerical outcomes, qualitative descriptions reflect the direction and significance of findings as reported. Heterogeneity in reporting practices across included studies precluded uniform quantitative synthesis and is acknowledged as a limitation of the present review (). Participant Characteristics: The 37 studies encompassed 8,432 total participants across all groups. Sample sizes ranged from 25 participants (minimum inclusion criterion) to 4,568 participants. The median study sample size was 89 participants (interquartile range: 38-247). Follow-up durations for observational longitudinal studies varied considerably, ranging from 6 months to 25 years, with a median follow-up of 5.2 years. Randomized controlled trials employed shorter intervention periods (4 weeks to 18 months) appropriate to experimental designs examining acute effects of abstinence or treatment.

Geographic and Temporal Distribution: Studies were conducted across multiple regions: North America (n=19, 51.4%), Europe (n=12, 32.4%), Australia/New Zealand (n=5, 13.5%), and Asia (n=1, 2.7%). Publication years spanned from 2000 to 2024, reflecting the evolution of neuroimaging methodologies and increasing focus on developmental neurobiology (). Cannabis Use Assessment Methods: Cannabis use patterns were assessed through structured clinical interviews in 28 studies (75.7%) and validated questionnaires in 9 studies (24.3%). Biochemical verification of recent use or abstinence was employed in 22 studies (59.5%) using urine or blood cannabinoid testing, while 15 studies (40.5%) relied primarily on self-report measures with clinical validation (, ).

Neuroimaging studies consistently demonstrate that timing of cannabis initiation critically influences brain structural development. Wilson et al. () established that individuals beginning use before age 17 showed smaller whole brain volumes and reduced cortical gray matter compared to later-onset users. Jacobus et al. () revealed paradoxically thicker cortical estimates across 23 brain regions in heavy adolescent users over three years, suggesting disrupted neuromaturation. Albaugh et al.’s () IMAGEN study of 799 initially cannabis-naive adolescents demonstrated dose-dependent accelerated cortical thinning in prefrontal regions over five years, with effects in cannabinoid receptor-rich areas linked to increased attentional impulsiveness. Batalla et al. () found gene-environment interactions between DAT1 variants and cannabis exposure affecting hippocampal volumes in early-onset users.

Yücel et al. () documented substantial hippocampal (12.0%) and amygdala (7.1%) volume reductions in heavy users, correlating with cumulative exposure and subthreshold psychotic symptoms. Zalesky et al. () demonstrated impaired connectivity in the fornix and corpus callosum correlating with age of initiation, while Shollenbarger et al. () revealed FAAH genetic variants modulate cannabis effects on frontolimbic white matter. Bava et al. () found that fractional anisotropy patterns in substance-using adolescents correlated differently with cognitive performance, suggesting altered neurodevelopmental trajectories.

Functional neuroimaging reveals cannabis effects alter brain network dynamics beyond structural changes. Orr et al. () found cannabis-dependent adolescents showed increased low-frequency oscillations and stronger intra-hemispheric but reduced interhemispheric connectivity, potentially reflecting compensatory mechanisms. Camchong et al. () demonstrated that while controls showed increasing functional connectivity between anterior cingulate cortex and superior frontal gyrus, cannabis use disorder participants exhibited decreasing connectivity between caudal ACC and dorsolateral/orbitofrontal cortices over 18 months. Lower baseline connectivity predicted greater subsequent cannabis use, with high use predicting lower IQ and slower cognitive function. Task-based fMRI studies reveal functional consequences during cognitive performance. Jager et al. () found adolescent cannabis users achieved normal working memory performance but required excessive prefrontal activation, suggesting inefficient neural recruitment. Tapert et al. () demonstrated that after 28 days abstinence, adolescent users showed significantly increased activation during inhibitory control tasks despite normal performance, indicating persistent neuroadaptive changes requiring greater processing effort ().

The cognitive impact of adolescent cannabis use presents a complex picture. Meier et al. (), following 1,037 participants from birth to age 38, revealed persistent neuropsychological decline in adolescent-onset users, with impairment concentrated among those with persistent use. Critically, cessation did not fully restore functioning, suggesting potentially lasting neurotoxic effects. However, this interpretation has been challenged by twin studies controlling for familial confounding. Jackson et al. () found that marijuana-using twins did not demonstrate significantly greater IQ decline than their abstinent siblings, suggesting observed deficits may reflect shared genetic and environmental factors rather than direct cannabis effects. Castellanos-Ryan et al. () found that poor working memory but high verbal IQ predicted earlier cannabis onset, while earlier onset and higher frequency were associated with verbal IQ decline and impaired trial-and-error learning independent of educational attainment.

Variability in cognitive findings becomes apparent in well-controlled studies. Mokrysz et al. () examining 2,235 teenagers, found no IQ or educational differences between heavy cannabis users and never-users after controlling for confounds, particularly cigarette smoking. Scott et al. () found complex patterns in 4,568 adolescents: while frequent users showed worse executive control, occasional users unexpectedly exhibited better performance than non-users, suggesting nuanced relationships that may reflect selection effects or resilience factors rather than simple dose-response relationships.

The reversibility of cognitive effects remains critical. Pope et al. (), comparing early-onset users (<17 years) with late-onset users after 28 days abstinence, found early-onset users performed worse on multiple measures, particularly verbal IQ. However, when controlling for verbal IQ, virtually all cognitive differences disappeared, suggesting effects may reflect pre-existing differences or educational disruption. Direct evidence for cognitive recovery came from Schuster et al.’s () randomized controlled trial, where 88 young regular users were assigned to 4 weeks of biochemically-verified abstinence or continued use. Results revealed significant improvements in verbal memory, driven primarily by gains during the first week of abstinence, providing experimental evidence that cognitive improvements can occur rapidly following cessation in young users.