Altered Neurobehavioral Reward Response Predicts Psychotic-Like Experiences in Youth Exposed to Cannabis Prenatally

Abstract

BACKGROUND:

Rates of prenatal cannabis exposure (PCE) are rising with increasingly permissive legislation, which may be a risk factor for psychosis. Disrupted reward-related neural circuitry may underlie this relationship. We aim to elucidate neural mechanisms involved in the association between PCE and youth-onset psychotic-like experiences (PLEs) by probing correlates of reward anticipation, a neurobehavioral marker of endocannabinoid-mediated dopaminergic function.

METHODS:

This longitudinal, prospective study analyzed task-related functional neuroimaging data from baseline (N = 11,368), 2-year follow-up (n = 7928), and 4-year follow-up (n = 2982) of the ongoing ABCD (Adolescent Brain Cognitive Development) Study, which recruited children ages 9 to 10 years at baseline from 22 sites across the United States.

RESULTS:

PCE (652 exposed youths) was longitudinally associated with PLEs. Blunted neural response to reward anticipation was associated with PLEs, with stronger effects observed in PCE youths (all |βs| > 0.5; false discovery rate [FDR]–corrected p < .05). This baseline hypoactivation predicted PLEs in middle adolescence (β = −0.004; p_FDR< .05) and mediated the relationship between PCE and PLEs. Dampened behavioral reward sensitivity was associated with PLEs across visits (|β| = 0.21; p_FDR < .001). PLEs were positively associated with trait-level measures of reward motivation and impulsivity, with stronger effects for PCE youths (all |βs| > 0.1; all p_FDR < .05).

CONCLUSIONS:

Blunted striatal activation may serve as a biomarker for disrupted reward processing and increased psychosis risk during development. PCE may affect childhood behaviors and traits related to altered reward sensitivity.

Past-month cannabis use among pregnant people increased by over 130% from 2003 to 2022 in the United States with increasingly permissive legislation (1,2). While there is growing evidence for associations between early cannabis use and elevated psychosis risk, less is known about the effects of cannabis exposure in utero (3–7). Recently, prenatal cannabis exposure (PCE) has been linked to psychotic-like experiences (PLEs) in early childhood (8,9). While the neurobiology that underlies the link between cannabis and psychosis is not fully understood, the brain’s reward pathways including the dopaminergic mesocorticolimbic pathway are implicated in both cannabis use disorder (CUD) and psychosis (10,11). It is unclear whether these pathways mechanistically link early cannabis exposure and psychosis proneness.

Cross-sectional evidence in a large sample of youth (N = 11,489) suggests that PCE increases risk for PLEs—subthreshold psychotic symptoms that precede and may predict the onset of overt psychotic disorder (12–14)—even in preadolescence (8). Resting-state functional connectivity (rsFC) studies have found that rsFC in the salience network mediates the association between PCE and child psychopathology in preadolescence (15), and PCE after maternal knowledge of pregnancy is associated with increased rsFC of the left putamen and the auditory network in 9-to 12-year-old children in the ABCD (Adolescent Brain Cognitive Development) Study cohort (16). However, the functional neural correlates of the relationship between PCE and PLEs have not yet been fully elucidated, nor have these relationships been modeled longitudinally. Furthermore, behavioral and trait-level factors that contribute to psychosis and substance use vulnerability including reward motivation and impulsivity have not yet been examined. There is evidence for diminished or disrupted reward processing (e.g., reduced responsiveness to reward) in both individuals with psychotic disorder and individuals at clinical high risk for psychosis (17,18). Furthermore, PLEs may be linked to altered reward responsivity even in the absence of full-blown psychosis (19–21). Impulsivity, which has been linked to cognitive control deficits, emotional dysregulation, and risk-taking behaviors, may increase vulnerability to PLEs by contributing to distortions in perception and belief formation (22). Furthermore, studying the relationship between PCE, PLEs, and their reward-related neural and behavioral correlates is of particular clinical significance in adolescence, which represents a crucial risk period for the onset of overt psychotic disorder and substance use vulnerability (3,6,16).

We examined the relationship between neurobehavioral markers of reward anticipation in task-based functional magnetic resonance imaging (fMRI) from the ABCD Study, a population-based, longitudinal cohort of 11, 878 children ages 9 to 10 years at baseline. We analyzed the reward anticipation phase of the Monetary Incentive Delay (MID) task, a task designed to probe reward response, as dopamine neurons in reward-related brain regions fire more readily in response to reward-predictive cues rather than reward delivery (23). This preferential firing during reward anticipation is thought to represent the motivational processing of incentive, attributed to reward mediated by larger systems encompassing mesolimbic dopaminergic transmission (24). Furthermore, hypoactivation during the anticipation phase of the MID task has been demonstrated in both adults with schizophrenia (24) and adults with CUD (25).

First, we aimed to test whether the association between PLEs and PCE persists into later adolescence. Next, the main hypotheses tested using baseline data (N = 11,368) are that 1a) PLEs would be associated with blunted neural activation in reward-related brain regions (i.e., the striatum and ventromedial prefrontal cortex [vmPFC]) in response to reward expectancy, and 1b) within these models, relationships would be stronger for PCE youths. Then, we examined the prospective relationship between brain activation to reward-predictive cues at baseline and PLEs at the 4-year follow-up (middle adolescence, n = 2982). We hypothesized that 2) blunted activation in reward-related brain regions in PCE youths at baseline in response to reward-predictive cues would be associated with more severe PLEs in middle adolescence. Third, we examined associations between PLEs by PCE status and behavioral and trait-level measures of reward motivation and impulsivity across time points, hypothesizing that 3a) behavioral response time to reward-predictive cues would be associated with PLEs and would differ by PCE status and 3b) that PLEs would be associated with greater trait-level reward responsiveness, drive, and impulsivity, with stronger effects for PCE youths.

METHODS AND MATERIALS

Data Source and Participants

Data came from children (N = 11,368 at baseline; mean ± SD age = 9.83 ± 0.6 years; 47.85% girls; 74.13% White) born between 2005 and 2009, in the ongoing longitudinal ABCD Study (data release 5.1.0; https://abcdstudy.org/). All parents/caregivers and children provided written informed consent and verbal assent, respectively, to a research protocol approved by the institutional review board at each data collection site throughout the United States (https://abcdstudy.org/studysites/) (Figure S1 and Supplement, section 1). This included data from 3 study time points: baseline, 2-year study follow-up visit, and 4-year study follow-up visit. For our analyses, data from participants with information on PCE who passed functional neuroimaging MID task quality control recommendations were included (see https://docs.abcdstudy.org/latest/documentation/imaging/#quality-control-and-recommended-image-inclusion-criteria).

Outcomes and Measures

Psychotic-Like Experiences.

PLEs were assessed using a summary score for level of distress from the Prodromal Questionnaire-Brief Child Version (PQ-BC) (26), a 21-item, developmentally appropriate questionnaire that was validated in the ABCD Study sample (for more information, see the Supplement, section 2).

Prenatal Cannabis Exposure.

PCE is based on parent or caregiver (9709 of 10,716 biological mothers [90.60%]) retrospective report. Three mutually exclusive groups were formed (at baseline): no exposure (n = 10,716), exposure prior to maternal knowledge of pregnancy only (n = 420), and exposure after maternal knowledge of pregnancy (this group includes exposure before and after knowledge of pregnancy; n = 232). ABCD Study prevalence estimates of self-reported prenatal cannabis use are consistent with toxicology-based prevalence estimates from national datasets collected during the years when these children were born, supporting the accuracy of self-report (27).

fMRI Paradigm.

ABCD imaging data collection, acquisition, and analysis have been described previously (28–30). The MID task is a well-validated probe designed to measure domains of reward processing, including anticipation and receipt of reward and modulation of motivation (31,32). See the Supplement, sections 1 and 3, and Figure S2 for details on task design and data acquisition and analysis. We assessed neural activation in the striatum and vmPFC during reward anticipation (Supplement, section 4).

Statistical Analyses

All models included age, biological sex, parental education, and income as covariates, with research site and family unit (nested within site to account for twins and triplets in this study sample) as random effects. To determine whether differences in PLEs between groups could be attributed to differences in other prenatal or environmental exposures (33), we included variables measuring birth weight, prenatal exposure to tobacco or alcohol (before or after maternal knowledge of pregnancy), unplanned pregnancy, and prenatal vitamin use in models testing group differences in the association between PCE and PLEs. Imaging analyses also included mean framewise displacement as a covariate (34). Models testing the association between baseline measures and PLEs at the 4-year follow-up included baseline PLEs as an additional covariate. Participant ID was included as a random effects term in longitudinal models. All results reported were false discovery rate (FDR) corrected for multiple comparisons. The PQ-BC data are zero inflated and positively skewed; therefore, models with PLEs as the dependent variable (DV) utilized Poisson distribution general linear models (log link function) (35). A q value < .05 was considered statistically significant. Results are expressed as standardized coefficients with 95% CIs. Figure S1 presents a schematic representation of data used across visits.

To address the question of whether associations between PCE and PLEs persist through middle adolescence, we conducted linear mixed model analyses with PLEs as the DV and cannabis exposure group (unexposed vs. exposed, with unexposed as the reference group) as the independent variable longitudinally across 3 waves of data. To test for differences regarding whether a child was exposed to cannabis before or after parental knowledge of pregnancy, as in Paul et al. (8), we reran the same models treating the independent variable as a 3-level factor (exposed after parental knowledge of pregnancy vs. exposed before parental knowledge of pregnancy vs. unexposed). Based on findings from this initial analysis that indicated that there was no statistically significant difference in PLEs between the 2 exposure groups, we collapsed across exposure groups for all subsequent analyses.

Cross-sectional relationships between activation in reward-related brain regions during reward anticipation and PLEs at baseline tested hypotheses that PLEs (DV) would be associated with 1) blunted activation in response to reward anticipation in a priori reward-related regions of interest (ROIs) and that 2) effects would be stronger for PCE youths. Linear mixed models tested the associations between activation in a priori ROIs and PLEs, with group × region interactions tested on each ROI (striatum and vmPFC), during reward anticipation (Supplement, sections 1 and 4). We also tested whether striatal activation mediated the relationship between PCE and PLEs by conducting a causal mediation analysis using the mediation package in R. As a control analysis to test the specificity of results to the dopaminergic mesocorticolimbic pathway, we reran these models with amygdala and insula ROIs; although not primary elements of this pathway, both regions are involved in the mechanisms of addiction, substance use, and reward-related decision making (36–39) (Supplement, section 5). We hypothesized that activation in the amygdala and insula would not be associated with PLEs.

To test the hypothesis that hypoactivation in reward-related brain regions at baseline would be prospectively associated with greater PLEs in adolescence, linear mixed models with PLEs at the 4-year follow-up as the DV and baseline activation in reward-related ROIs as independent variables were tested across groups. We utilized baseline neuroimaging data to establish the predictive validity of an early biomarker of elevated risk for developing PLEs, potentially present as early as childhood.

We analyzed behavioral performance on the MID task (reaction time [RT] to reward-predictive cues) using longitudinal linear mixed-effects models, with PLEs as the DV and RT as the independent variable to test the hypothesis that behavioral responses to reward-predictive cues would differ by PCE status (see Supplement, section 1 for more detail on task conditions). Then, we tested group × behavior interactions to test the hypothesis that the relationship between behavioral performance and PLEs would differ as a function of exposure status. Behavioral models were tested across all time points to characterize behavioral differences regardless of developmental period.

Then, we investigated the relationship between PLEs and trait-level factors related to reward motivation and impulsivity by PCE status to test the hypothesis that PLEs would be associated with these trait-level factors, with stronger effects for PCE youths. Specifically, self-report trait measures included reward responsiveness and drive, as measured by the Behavioral Inhibition System and Behavioral Activation System (BIS/BAS) scales (40,41); excitement about rewards, as measured by a post-MID task questionnaire; and 4 measures of impulsivity, including 1) negative urgency, 2) lack of planning, 3) sensation seeking, and 4) positive urgency as measured by the UPPS-P Impulsive Behavior Scale (42). These self-report measures were collected at baseline, the 2-year follow-up, and the 4-year follow-up; all 3 time points were included in our analyses. Longitudinal linear mixed models tested the associations between trait measures and PLEs, with group × trait interactions tested on each trait measure. See Supplement, section 6 for information on trait measures.

RESULTS

Participant Demographics

We analyzed task-based fMRI and behavioral data acquired during the MID task from participants in the ABCD Study (https://abcdstudy.org) at baseline (N = 11,368; mean age, years = 9.83 ± 0.6; 47.85% girls), the 2-year follow-up (7479 unexposed youths; 144 exposed after knowledge of pregnancy; 305 exposed before knowledge of pregnancy), and the 4-year follow-up (2826 unexposed youths; 61 exposed after knowledge of pregnancy; 95 exposed before knowledge of pregnancy). Table 1 describes baseline demographics of our sample stratified by PCE status. Due to the ABCD Study data release schedule, ABCD Data Release 5.1.0 includes only a subset of the full sample expected for year 4 data (approximately 50% of the expected data for this time point). Table S1 includes details on rates of participant substance use by study time point.

Table 1.

ABCD Study Sample Characteristics at Baseline

Variable	PCE Post Knowledge of Pregnancy, n = 232	PCE Before Knowledge of Pregnancy, n = 420	No PCE, n = 10,716
Child Variables
Age at baseline, months	117.90 (7.44)	119.05 (7.70)	119.05 (7.49)
Sex, female	126 (54.3%)	199 (47.4%)	5123 (47.8%)
Birth weight, kg	6.36 (1.53)	6.63 (1.39)	6.63 (1.39)
Baseline FD during MID, mma	0.42 (0.45)	0.37 (0.40)	0.31 (0.36)
Year 2 FD during MID, mmb	0.29 (0.37)	0.31 (0.35)	0.25 (0.29)
Year 4 FD during MID, mmc	0.23 (0.34)	0.19 (0.17)	0.16 (0.19)
Race/Ethnicity
Asian	3 (1.24%)	3 (0.73%)	193 (1.78%)
Black	105 (43.39%)	170 (41.16%)	2089 (19.28%)
Hispanic	33 (14.2%)	90 (21.5%)	2191 (20.5%)
Native American	14 (5.79%)	24 (5.81%)	351 (3.24%)
Other	9 (3.72%)	33 (7.99%)	733 (6.77%)
Pacific Islander	0 (0.00%)	4 (0.97%)	34 (0.31%)
Whited	144 (59.50%)	242 (58.60%)	8203 (75.72%)
Pregnancy and Family Variables
Unplanned pregnancya	164 (70.69%)	304 (72.38%)	3837 (35.81%)
Prenatal vitamin usee	171 (73.7%)	379 (90.2%)	9872 (92.2%)
Household Incomed
$0–$49,999	181 (78.02%)	307 (73.10%)	6267 (58.48%)
$50,000–$74,999	16 (6.90%)	50 (11.90%)	1577 (14.72%)
$75,000–$99,999	11 (4.74%)	23 (5.48%)	859 (8.02%)
$100,000–$199,999	4 (1.72%)	12 (2.86%)	836 (7.80%)
≥$200,000	8 (3.45%)	3 (0.71%)	186 (1.74%)

Values are presented as mean (SD) or n (%).

ABCD, Adolescent Brain Cognitive Development; FD, framewise displacement; MID, Monetary Incentive Delay; PCE, prenatal cannabis exposure.

^aNo PCE, PCE both before and post parental knowledge of pregnancy.

^bPCE before parental knowledge of pregnancy > both no PCE and PCE post parental knowledge of pregnancy.

^cPCE post parental knowledge of pregnancy > both no PCE and PCE before parental knowledge of pregnancy.

^dNo PCE > PCE both before and post parental knowledge of pregnancy.

^ePCE post parental knowledge of pregnancy < both no PCE and PCE before parental knowledge of pregnancy.

Longitudinal Associations Between PCE Status and PLEs

PCE was associated with more severe PLEs longitudinally across 3 waves of data relative to no exposure (see Table 2 for full results). Analyses assessing differences in youths exposed to cannabis before versus after parental knowledge of pregnancy revealed that cannabis exposure both before and after parental knowledge of pregnancy was associated with more severe PLEs compared with no exposure, with no statistically significant differences between the 2 exposure groups.

Table 2.

Group Differences in the Longitudinal Associations Between PCE and PLE

	Exposed vs. Unexposed, Across Exposure Status Groups		PCE-A vs. Unexposed		PCE-B vs. Unexposed		PCE-A vs. PCE-B
	Standardized β	Standardized CI	Standardized β	Standardized CI	Standardized β	Standardized CI	Standardized β	Standardized CI
PLE	0.370***	0.21 to 0.52	0.470***	0.22 to 0.73	0.260**	0.07 to 0.45	0.286	0.00 to 0.08
PLE in Subset of Neuroimaging Participants	0.339***	0.04 to 0.11	0.292**	0.01 to 0.07	0.166*	0.00 to 0.06	0.165	−0.01 to 0.05

PCE was significantly associated with PLEs, across all time points, both in the full sample and in the subset of participants for whom Monetary Incentive Delay task data and 4-year follow-up data were available. There were no statistically significant differences in PLEs between PCE-A and PCE-B groups. Covariates: age, biological sex, parental education, income, birth weight, prenatal exposure to tobacco or alcohol before or after maternal knowledge of pregnancy, unplanned pregnancy, and prenatal vitamin use, with research and family unit (nested within site to account for twins and triplets in this study sample) as random effects.

^*q < .05,

^**q < .01,

^***q < .001, false discovery rate corrected.

PCE, prenatal cannabis exposure; PCE-A, PCE after parental knowledge of pregnancy; PCE-B, PCE before parental knowledge of pregnancy; PLE, psychotic-like experience.

Baseline Associations Between ROI Activation and PLEs by PCE Status

Figure S3 shows the results of models testing the relationships between PCE, activation in a priori ROIs, and PLEs in youths. As hypothesized, striatal activation during reward anticipation was inversely associated with reported PLEs and was blunted in PCE youths compared with unexposed youths (Table 3). While there was no significant association between vmPFC activation and PLEs, vmPFC activation was blunted in PCE youths compared with unexposed youths. An exploratory analysis examining the large loss condition revealed a similar pattern (Supplement, section 7). Results of secondary analyses revealed that PLEs were not associated with amygdala or insula activation during reward anticipation (Table 3 and Figure S6).

Table 3.

Group Differences in Baseline Association Between Neural Activation to Large Reward Anticipation Cues in the MID Task and PLEs and Longitudinal Association Between Behavioral Response to Large Reward in the MID Task and PLEs

Group Differences in the Baseline Association Between Neural Activation to Large Reward in the MID Task and PLEsa
	Effect of PLEs		Effect of Group, Exposed vs. Unexposed		Group × Region Interaction
	Standardized β	Standardized CI	Standardized β	Standardized CI	Standardized β	Standardized CI
Striatum	−0.050*	−0.10 to −0.01	0.060**	0.02 to 0.11	−0.005	−0.14 to 0.21
vmPFC	−0.020	−0.07 to 0.02	0.070**	0.03 to 0.11	0.004	−0.17 to 0.17
Amygdala	−0.052	−0.05 to 0.02	0.227**	0.02 to 0.09	0.088	0.00 to 0.06
Insula	0.008	−0.03 to 0.03	0.200*	0.01 to 0.08	−0.078	−0.04 to 0.03
Group Differences in the Longitudinal Association Between RT to Large Reward in the MID Task and PLEsb
	Psychotic-Like Experiences		Effect of Group, Exposed vs. Unexposed		Group × RT Interaction
RT, ms	0.210***	0.18 to 0.23	0.09	−0.06 to 0.25	0.25	−0.08 to 0.20

MID, Monetary Incentive Delay; PCE, prenatal cannabis exposure; PLE, psychotic-like experience; ROI, region of interest; RT, reaction time; vmPFC, ventromedial prefrontal cortex.

^*q < .05,

^**q < .01,

^***q < .001, false discovery rate corrected.

^aStriatal activation to reward-predictive cues during the MID task was inversely associated with reported PLEs. However, there was no significant association between vmPFC activation and PLEs. There were significant main effects of group; PCE youths showed more blunted activation in both the striatum and vmPFC during reward anticipation compared with unexposed youths. There were no significant group × region interactions. Secondary analyses examining amygdala and insula ROIs revealed that there were no associations with PLEs. PCE youths showed more blunted activation in both the amygdala and insula during reward anticipation compared with unexposed youths. Reference level for group effect: youths exposed to cannabis prenatally. Covariates: age, biological sex, parental education, income, and mean framewise displacement as fixed effects, with research and family unit (nested within site to account for twins and triplets in this study sample) as random effects.

^bAssociations between PLEs and RT in the MID task by group status (exposed to cannabis prenatally vs. unexposed) to high-magnitude reward-predictive cues. RT was positively associated with PLEs in both groups. Covariates: age, biological sex, parental education, and income as fixed effects with research and family unit (nested within site to account for twins and triplets in this study sample) as random effects.

Mediation Effect

Striatal activation was a significant mediator of the relationship between PCE and PLEs at baseline. The average causal mediation effect was 0.0130; 95% CI, 0.001–0.03; p = .04, indicating a small but significant indirect effect. The average direct effect remained significant (0.460; 95% CI, 0.252–0.66; p < .001), with a total effect of 0.473; 95% CI, 0.265–0.67; p < .001. The estimate of the total mediation effect was 0.027 (95% CI, 0.001–0.07). The mediation effect is depicted in Figure 1.

Figure 1.

Mediation model examining the relationship between prenatal cannabis exposure (PCE), blunted neural response to reward anticipation, and psychotic-like experiences (PLEs). PCE is significantly associated with blunted neural response to reward anticipation (β = 0.460; 95% CI, 0.252–0.66; A path). In turn, blunted neural response to reward anticipation predicts increased PLEs (β = 0.013; 95% CI, 0.0008–0.03; B path). The total effect of PCE on PLEs (β = 0.474; 95% CI, 0.265–0.67; C path) remains significant after accounting for the mediator, with the direct effect (β = 0.460; 95% CI, 0.252–0.66; C’ path) indicating that neural hypoactivation partially mediates this relationship. The proportion of the total effect mediated by blunted neural response to reward anticipation is 2.7%. *p < .05, ***p < .001.

Association of Baseline ROI Activation With PLEs at the 4-Year Follow-Up Visit by PCE Status

Figure 2 shows results of models using baseline fMRI data to predict PLEs at the 4-year follow-up. As hypothesized, activation in the striatum during reward anticipation at baseline was associated with greater PLEs at the 4-year follow-up across groups, despite controlling for baseline PLEs. However, consistent with cross-sectional findings, baseline vmPFC activation was not significantly associated with PLEs.

Figure 2.

(A) Regions of interest (ROIs) in the current study included the striatum (blue) and ventromedial prefrontal cortex (vmPFC) (red). Results of ROI analyses testing prospective associations between baseline activation during reward anticipation in (B) the striatum and (C) the vmPFC on the Monetary Incentive Delay task across groups and psychotic-like experiences (PLEs) at year 4. Striatal activation was inversely associated with PLEs at year 4 (B), whereas there was no significant association between vmPFC activation to reward anticipation at baseline and PLEs at year 4 (C). Predicted values for youths who were not exposed to cannabis prenatally are depicted in teal, and predicted values for youths who were exposed to cannabis prenatally are depicted in purple. Shading around the best-fit lines indicates standard error. Covariates: age, biological sex, parental education, income, mean framewise displacement, and baseline PLEs as fixed effects with research and family unit (nested within site to account for twins and triplets in this study sample) as random effects. *p < .05, false discovery rate corrected. PQ-BC, Prodromal Questionnaire-Brief Child Version.

Longitudinal Associations Between PLEs and Behavioral Measures by PCE Status

Models testing the association between PLEs and behavioral performance on the MID task revealed that RT to reward-predictive cues was positively associated with PLEs in both exposure groups (Figure S4 and Table 3).

Longitudinal Associations Between PLEs and Trait Measures by PCE Status

PLEs in PCE youths were associated with greater reward responsiveness, drive, excitement about reward receipt, negative urgency, and all 4 domains of impulsivity compared with unexposed youths, across time points (Tables S2 and S5). PCE youths scored higher across all trait-level measures. There were significant group × trait interactions for drive, negative urgency, lack of planning, and sensation seeking, such that relationships to PLEs were stronger for PCE youths.

DISCUSSION

This study is the first to examine longitudinal relationships between PCE, PLEs, and behavioral and neural indices of reward anticipation over time in youth. First, we found that previously observed associations between PCE and PLEs in childhood persisted through middle adolescence. Second, this study offers novel evidence implicating the brain’s reward pathways in the link between PCE and early signs of psychosis. Specifically, 1) blunted neural response to reward anticipation was associated with PLEs, with stronger effects observed in PCE youth; 2) striatal hypoactivation at baseline predicted PLEs in middle adolescence; 3) dampened behavioral reward sensitivity was associated with PLEs over time; and 4) PLEs were positively associated with trait-level measures of reward motivation and impulsivity across multiple time points, with stronger effects for PCE youths.

Our study extends recent work demonstrating an association between PCE and PLEs in 9- and 10-year-old children exposed to cannabis after parental knowledge of pregnancy, which remains robust through middle adolescence after rigorously controlling for covariates (8). Not only did the association persist through middle adolescence, but the longitudinal association also held regardless of whether a child was exposed before or after parental knowledge of pregnancy, indicating that even early exposure contributes to psychotic-like symptomatology later in childhood. The cannabinoid CB₁ receptor (CB1R) is not expressed in the fetus until 5 to 6 weeks gestation, which corresponds approximately with knowledge of pregnancy (43–45); our findings that there were no statistically significant differences in PLEs between exposure groups support hypotheses suggesting that there may still be indirect associations through endocannabinoid receptor expression in the placenta (46). We cannot rule out that other parental characteristics that underlie cannabis use, such as personality traits, may confound this relationship. More granular studies of dose, frequency, and gestational age of exposure are crucial for a better understanding of cannabis’ impact on the developing brain.

In addition, we provide the first evidence that hypoactivation in the striatum during anticipation of reward at ages 9 to 10 is associated with PLEs, with stronger effects for PCE youths; this hypoactivation also predicts PLEs in middle adolescence. PCE youths also showed more blunted activation in the vmPFC during reward anticipation compared with unexposed youths. Moreover, we found that striatal hypoactivation mediated the relationship between PCE and PLEs. Preclinical research shows adolescent and adult animals prenatally exposed to Δ⁹-tetrahydrocannabinol (THC) and other CB1R agonists exhibit alterations in reward-related neural circuitry (7). Our results are consistent with previous work demonstrating hypoactivation of the striatum during reward anticipation in adult patients with psychosis (47), extending this work to a youth sample with subthreshold psychotic symptoms. Further consistent with these results, we observed evidence that striatal hypoactivation mediated the relationship between PCE and PLEs. We also extended a previous study that reported blunted activation of the striatum to monetary reward anticipation in adult cannabis users (25), which may imply diminished orientation to a future reward. The PFC and value-based decision-making functions are not yet fully developed in this age group (48–52), which may explain the specificity of the association between PLEs and blunted striatal activation, versus no association with vmPFC activation. Because the vmPFC matures through adolescence and into early adulthood, it is likely that its functional contributions to reward processing and regulation are not yet fully established in this age group, which may limit its predictive role. This finding is also consistent with evidence that striatal dysfunction is a core feature of psychosis risk, given the region’s dopamine receptor expression and role in reward processing (10). The fact that there were no associations between PLEs and amygdala or insular activation implies specificity of these effects to the mesocorticolimbic pathway and striatal involvement in reward-related deficits linked to PLEs in this age group rather than broader affective or interoceptive dysfunction (53). These findings also extend recent work showing that PCE is longitudinally associated with greater striatal rsFC in children ages 9 to 12 years (54), consistent with evidence of altered connectivity between the cortex and striatum during acute cannabinoid administration (43). PCE may interfere with the development of frontostriatal circuits critical for flexible and adaptive responses to motivational cues, whether appetitive or aversive, thereby increasing vulnerability to PLEs. Future work should elucidate whether the changing strength of frontostriatal connectivity during neurodevelopment is associated with youth psychopathology and/or a domain-general deficit in salience processing.

The endocannabinoid system, specifically the CB1R, is the site of action of THC, the primary psychoactive component of cannabis. CB1Rs are preferentially expressed in key reward-related brain areas and regulate dopaminergic signaling (48–50). While a causal mechanism that links PCE to later onset of PLEs has not yet been established, early alterations to the endocannabinoid-mediated dopaminergic system may contribute to vulnerability to psychosis in PCE youth (51–53,55–57). Our finding of prospective associations between baseline activation in reward-related brain regions and increased PLEs in adolescence supports one potential mechanism through which PCE may lead to enhanced vulnerability to psychosis and produce long-term alterations in neural circuits involved in reward processing implicated in both acute PLEs and increased risk for the development of overt psychotic disorder. Moreover, blunted striatal activation may represent a biomarker of disrupted reward processing associated with subsequent PLEs, which is observable as early as ages 9 and 10. Rates of youth substance use are low at the early study time points currently available; therefore, this study is underpowered to test for potential effects of youth substance use. Future studies are warranted to investigate whether this blunted striatal activation in childhood is also associated with later risk for substance use in adolescence. Future studies should also investigate whether PCE is associated with greater self-reported adolescent substance use.

The observed association between slower RTs to reward-predictive cues and PLEs in the current study may suggest dampened sensitivity and/or incentive to reward-related cues. RT to reward-related cues is thought to represent a behavioral correlate of sensitivity and incentive, with faster RTs to reward representing greater reward-related motivation (58). Other studies have found that compared with typically developing youth, youth at clinical high risk for psychosis display impaired behavioral performance during reward and cognitive tasks, associated with altered activation in reward-related brain regions and greater psychosis symptom severity, regardless of cannabis use history (59–61). Early THC exposure in both rodents and humans results in behavioral changes relevant to reward processing during adulthood such as increased risky decision making and impulsivity and increased drug self-administration in rodents (62–66). While we observed dampened behavioral and neural response to reward in the current study, we also observed heightened trait-level sensitivity to reward. While, to our knowledge, this has not previously been studied in relation to PCE, traits including increased reward responsiveness and impulsivity have previously been linked to adult cannabis use (67). In a rat model, high-dose THC exposure in adolescence results in risky decision making and impulsivity in adulthood, as well as changes in reward learning, similar to findings seen in people who use cannabis (68). Although no causal model has been established, cannabis exposure may lead to neural habituation by which the brain becomes less responsive to reward cues, while higher trait-level reward sensitivity and sensation-seeking scores may instead reflect a tendency to seek out and value rewards but not necessarily heightened neural responsiveness to rewards. Self-reported excitement about rewards may be shaped by motivational tendencies, cognitive expectations, or compensatory mechanisms that operate independently of immediate neural responses (69–71).

Strengths and Limitations

This study leveraged the largest prospective neuroimaging and behavioral study of adolescence to date to examine a key environmental risk variable associated with increased risk for psychosis. Because the ABCD Study is an ongoing 10-year study, follow-up waves can be used to investigate whether childhood and adolescent PLEs and reward-processing deficits predict the persistence of psychopathology into later adolescence and young adulthood. Task-based functional imaging may offer a valuable biomarker for understanding mechanisms by which early cannabis exposure may contribute to later mental health outcomes.

This study also has certain limitations that should be noted. Self-reported cannabis use was reliant on maternal recall, an important study limitation. Nevertheless, it is important to note that prevalence estimates of self-reported prenatal cannabis use are consistent with toxicology-based prevalence estimates from national datasets collected during the years when these children were born, supporting the accuracy of the self-report (28). Furthermore, there is a lack of detailed information on cannabis potency, frequency of use, gestational age of exposure, and quantity of cannabis exposure during pregnancy in the ABCD Study. While we rigorously controlled for multiple demographic and other covariates, unmeasured confounders might have played a role. It is crucial to note that the small effect sizes commonly observed in ABCD studies of neural predictor variables should be taken into account when considering any neuroimaging metric as a biomarker of psychiatric outcomes. In addition, in future studies, the investigators may wish to further assess the utility of striatal hypoactivation as a potential biomarker by exploring the potential for individualized prediction of outcomes.

Conclusions

The results of this study demonstrate the significance of neurobehavioral markers of reward anticipation in the association between PCE and youth PLEs. Ultimately, we provide important new evidence for lasting effects of early cannabis exposure on reward-related neural circuitry that may be a pathway impacting early psychosis risk.

Supplementary Material

Supplementary material cited in this article is available online at https://doi.org/10.1016/j.biopsych.2025.05.019.