We found increased resting-state functional connectivity between the thalamus and the inferior and superior parietal cortices in particular as well as the superior and middle frontal cortices in adult cannabis non-users with a childhood diagnosis of ADHD.

The thalamus has been hypothesized to contribute to an increased addiction risk by enhancing responses to salient drug cues and attenuating the ability to inhibit these responses [21]. Salience processing refers to the identification of important external and/or interoceptive stimuli. The anterior insula and anterior midcingulate cortex are the central components of the brain’s salience network, and distributed brain regions, including the inferior parietal cortex and basal ventromedial nucleus of the thalamus, also participate in this large-scale brain network [35,36]. Our present findings of increased resting-state functional connectivity between the thalamus and inferior parietal cortex in cannabis non-users may thus be regarded in the context of salience processing toward drug cues. However, the inferior parietal lobe is also a convergence zone for different mental abilities, including language, attention, mathematical calculation, and social cognition [37]. Furthermore, long-term heavy cannabis use is associated with impaired working memory [38], which may be related to damage to the superior parietal cortex [39]. In one study, even after prolonged abstinence, brain processing efforts were greater in cannabis users than in non-users during inhibitory control, as evidenced by an increased blood-oxygen-level-dependent response in the frontal cortex as well as inferior and superior parietal cortices [40]. In summary, the increased thalamoparietal connectivity observed in cannabis non-users may reflect a better ability to control impulsive responses to drug cues [41], which would in turn facilitate more appropriate attribution of salience to behaviorally relevant stimuli other than cannabis (Table 4).

Comparison samples	Findings	Possible implications
Cannabis users and non-users with childhood ADHD	Increased thalamoparietal connectivity in non-users	Better ability to control impulsive responses to drug cues, accompanied by more appropriate attribution of salience to behaviorally relevant stimuli other than cannabis
Cannabis non-users with childhood ADHD and LNCG without childhood ADHD	Increased thalamoparietal connectivity in non-users	Possible improvement in the ability to control impulsive responses to drug cues, accompanied by more appropriate attribution of salience to behaviorally relevant stimuli other than cannabis
Cannabis non-users with childhood ADHD and LNCG without childhood ADHD	Increased thalamofrontal connectivity in non-users	Possible improvement in inhibitory control, which helps inhibit responses to drug cues and reduce the frequency of cannabis use
Cannabis users with childhood ADHD and LNCG without childhood ADHD	No significant difference in thalamic connectivity	N/A

The increased thalamoparietal connectivity in cannabis non-users with childhood ADHD was not only evident in the comparison with regular users of cannabis who had been diagnosed with ADHD but also in the comparison with the LNCG. Moreover, cannabis non-users with childhood ADHD showed increased thalamofrontal connectivity. The thalamus appears to mediate inhibitory control via its connectivity with the prefrontal cortex [42], and age-related progressive strengthening of thalamofrontal functional connectivity was observed in a sample of healthy children and adults [43]. Although patients with ADHD do not generally outgrow the disorder by adulthood, intermittent periods of remission can be expected in most cases [44]. Moreover, there has been a study showing that functional connectivity in the brains of adults whose inattentive symptoms had resolved did not differ significantly from that of their never-affected peers [45]. Therefore, adults who do not regularly use cannabis but have a past history of ADHD perhaps constitute a subgroup that provisionally attains relatively better thalamofrontal and thalamoparietal function, which can help inhibit responses to drug cues and thus reduce the frequency of cannabis use (Table 4).

Other brain regions that show reduced functional connectivity with the thalamus in cannabis users have also been implicated in addiction in the literature. Cannabis users demonstrated significantly less activation of the precentral and postcentral regions during a finger-sequencing task compared with non-users [46,47]. Furthermore, the persistence of ADHD was associated with smaller precentral and postcentral cortical thickness, while the early onset of cannabis use in patients with ADHD was associated with greater postcentral cortical thickness [48]. Altered activation of the primary somatosensory cortex was associated with risky decision-making in patients with substance use disorder [49]. We found large symmetric differences in thalamic connectivity in both hemispheres, which is consistent with the almost exact mirror symmetry of resting-state connectivity between the thalamus and cortex in healthy young adults [50].

Using the ACPI MTA data, Kelly and colleagues reported significant main effects of cannabis use history in two out of the nine large-scale intrinsic connectivity networks: the right superior temporal sulcus within the default network and the left fusiform gyrus within the lateral visual network both exhibited stronger intrinsic functional connectivity in cannabis users relative to non-users [29]. The nine large-scale intrinsic connectivity networks examined in this report included one network predominantly involving subcortical regions such as the amygdala, putamen, and thalamus; however, no significant findings were observed in the analysis using the predominantly subcortical network. The present work differs from that of Kelly and colleagues in that we specifically focused on the functional connectivity of the thalamus, based on a seed-to-whole brain approach using the thalamus as the seed, and found significant difference in thalamic connectivity between ADHD cannabis users and non-users. To further explore whether this difference was due to a decrease in connectivity among cannabis users or to an increase in connectivity among cannabis non-users, we performed additional analyses to compare their thalamic functional connectivity with that in the LNCG without childhood ADHD who were cannabis non-users. The only difference found in these subsequent analyses was increased bilateral thalamic functional connectivity in non-users with ADHD, and no significant difference in thalamic connectivity was observed between ADHD cannabis users and the LNCG. Albeit with different methods, Kelly and colleagues also did not observe a decrease in connectivity among cannabis users within the nine large-scale intrinsic connectivity networks tested, which is in line with the interpretation of our findings as an increase in connectivity among ADHD cannabis non-users rather than a decrease in connectivity among ADHD cannabis users.

This study has several limitations. In the original MTA, four different treatments were randomly assigned to children with ADHD. However, no information was available on which participant received which treatment, whether the participant continued treatment during a period of extended follow-up, or the presence or absence of ADHD at the time of scanning. For example, it would also be interesting to examine whether thalamocortical connectivity displays any variations according to the level of compliance with medication. In addition, no information was available on the type of stimulant medication used for each participant (e.g., methylphenidate or dextroamphetamine) in the ACPI database. Moreover, behavioral measures of ADHD symptoms and cannabis use were not included in the dataset. Therefore, we were not able to explore the functional implications of altered thalamocortical connectivity. Considering that the original MTA collected a comprehensive set of clinical information, integrating those data into the ACPI database would help further research. In particular, more information is needed about cannabis use, such as the type of cannabis used, route of administration, as well as date of last use. For example, we were not able to verify if any participant had recently consumed cannabis before imaging acquisition, which is an important limitation considering that cannabinoids may affect brain functioning for 28 days after last use [51]. In addition, information about cannabis use was collected based on self-report, limiting data reliability. Moreover, no information was available on other psychiatric or substance use disorders in the ACPI database. Considering that ADHD is frequently comorbid with other mental disorders, absence of such information is an important limitation of this study. Finally, the sample size was small, and we only included men in our analysis because of the small number of women, which limits the generalizability of our findings. However, it is noteworthy that a high male predominance is among the typical characteristics of the population of interest [1,30]. The exclusion of participants with excessive head motion resulted in a further reduction of the study sample size, but may also contribute to the strength of the study. As the current study was based on an existing dataset, we were unable to alter the sample size, and despite the small sample size, we considered that the ACPI dataset was worthy of research as it contains unique data collected over a long time frame that would allow investigation of the link between prior childhood ADHD and later marijuana use in adulthood. Notably, despite the small sample size, our findings were observed at a relatively strong statistical significance level in terms of cluster size and p-value. Many of the findings survived a stringent Bonferroni correction for two thalamic seeds tested as well as four different preprocessing pipelines (p < 0.00625; 0.05/8). We expect that the current findings will provide an initial basis for further studies with a larger collection of data.