Our main findings were that there was a bilateral increase in regional putamen shape bilaterally and in the right thalamus in PwP + C compared to PwP − C; whilst there was no evidence for selective regional shape deflation across the subcortical regions. There were no shape changes in controls with and without a history of cannabis use.

The putamen, alongside the caudate, together constitute the dorsal striatum which has been implicated in habit learning and compulsive behaviour (Everitt & Robbins, 2013). Increased putamen grey matter volume (Moreno‐Alcázar et al., 2018) and cerebral blood flow (Filbey, Aslan, Lu, & Peng, 2018) have been shown in healthy cannabis users compared to controls. Increased putamen volume has been shown to be associated with methamphetamine use (Andres et al., 2016; Chang et al., 2005; Jan, Lin, Miles, Kydd, & Russell, 2012), with alterations in putamen myeleoarchitecture in those at behavioural risk of addiction (Nord et al., 2019), and increased putamen volumes in probands with stimulant addiction and relatives of those with addictive disorder (Ersche et al., 2012). Two previous studies have examined cannabis use in the context of psychosis: one showed putamen volume increases in cannabis‐using compared to non‐using PwP (Koenders et al., 2015), whereas the other found neurochemical changes in the left putamen with higher NAA/choline ratio in cannabis‐using PwP compared to controls (Malchow et al., 2013). Our study thus extends their findings by demonstrating regional structural change using an alternative analytical strategy and confirming the finding of putamen enlargement in the PwP + C group.

We also find regional enlargement of the right thalamus in PwP + C compared to PwP − C. Consistent with our findings heavy cannabis users have been shown to have greater grey matter density in the right thalamus compared to non users (Matochik, Eldreth, Cadet, & Bolla, 2005), although volume decrease in cannabis users at risk for psychosis have also been reported (Kumra et al., 2012; Welch et al., 2011). The thalamus has a key role in the addiction circuitry and has been conceptualised to influence addiction pathology via cortico‐striatal‐thalamocortical loops (Huang, Mitchell, Haber, Alia‐Klein, & Goldstein, 2018). Thalamic nuclei have been demonstrated to regulate instrumental learning, behavioural flexibility and behavioural reinforcement (Bradfield, Hart, & Balleine, 2013; Lalive, Lien, Roseberry, Donahue, & Kreitzer, 2018; Marton, Seifikar, Luongo, Lee, & Sohal, 2018). Cortico–striatal–thalamocortical loops are particularly implicated in an array of psychopathology with compulsive behaviour including obsessive compulsive disorder and substance use (Fettes, Schulze, & Downar, 2017). In contemporary addiction models the dorsal striatal‐pallidal‐thalamic component of the circuit is particularly important in hardwiring habit formation the behavioural output driving compulsive use (Everitt & Robbins, 2005; Koob & Volkow, 2016). Taken together speculatively our findings of morphologically changes in the putamen and thalamus in PwP + C versus PwP – C, in substrates involved in habit formation, may represent a biomarker of compulsive use in this group. This may go some way to explaining why some PwP use cannabis despite clearly adverse consequences. In the absence of longitudinal analysis in this group, however, we are unable to disambiguate whether these changes were secondary to or pre‐existed the cannabis use.

Supporting this interpretation we show similar changes (bilateral putamen and right thalamic regional enlargements) in the whole sample in a secondary analysis when we reanalysed the data to compare lifetime cannabis dependence versus non dependence. However we are cautious of labelling these findings as a marker of substance dependence per se. This was not the main purpose of this study where we preferentially recruited a sample of PwP. We note a recent large study examining subcortical morphology found negative effects of alcohol dependence, positive effects of nicotine dependence and no main effect for other substance dependence (Chye et al., 2019b) and so limit our interpretations to those of a sample with early psychosis.

In PwP we find evidence that cannabis is associated with shape enlargement in structures involved with compulsive use. We do not find this in controls but do find increased volume (effect size 0.67) as compared to non‐users in the right nucleus accumbens (ventral striatum). In contemporary models both substrates serve different purposes in addiction circuits with the ventral striatum involved in mesolimbic reward and impulsive behaviour, and the dorsal striatum involved in habit formation and compulsive drug‐seeking (Everitt & Robbins, 2013; Koob & Volkow, 2010; Koob & Volkow, 2016). This raises the possibility of differing neurobiological alterations in cannabis‐using PwP versus otherwise healthy cannabis users. In our study the parameters of cannabis consumption including quantity, proportion of current users, and age of first use of cannabis did not significantly differ between cannabis using groups. Nevertheless, more individuals had a trend towards lifetime diagnosis of cannabis use disorder in the PwP + C group than the C + C group (79 vs. 50%). This is not unexpected given the association between cannabis dependence and psychosis (Fergusson, Horwood, & Swain‐Campbell, 2003; Wisdom, Manuel, & Drake, 2011) and may be a possible explanation for the neurobiological findings suggesting that the two cannabis using groups had been at differing stages of the addiction cycle. Different circuitry may explain why PwP use cannabis whilst in participants without psychotic disorder, psychotic experiences are associated with cessation of cannabis use (Sami, Notley, Kouimtsidis, Lynskey, & Bhattacharyya, 2019). In particular one could speculate that there is an increase in compulsivity in PwP compared to controls with a history of cannabis use, which may be associated with cannabis use disorder. Such a formulation could form the basis of hypothesis testing for further research though a multimodal approach.

We did not find evidence for regional deflation in cannabis‐using compared to non‐using PwP as we had particularly expected in the hippocampus. Reductions in volume and shape are considered to be consistent with an effect of regular cannabis use and are seen in chronic populations (Rais et al., 2008; Solowij et al., 2013). Longitudinal studies demonstrate that this reduction follows cannabis use (Rais et al., 2008; Rais et al., 2010; Welch et al., 2011; Welch, Stanfield, et al., 2011). This may support the view that the reduction in grey matter seen in long‐term users is a cumulative effect of cannabis use (Sami & Bhattacharyya, 2018).

There are several strengths in this study. We include four groups, including the control cannabis using group which has often been omitted from similar previous studies (Sami & Bhattacharyya, 2018). The T1‐weighted scans used are isotropic and relatively high resolution volumetric scans for addressing this question. We undertook segmentation using FSL FIRST which has been shown to be superior to FreeSurfer for putamen segmentation and comparable to gold standard expert manual tracing for basal ganglia structures (Perlaki et al., 2017). We covary for alcohol and tobacco use, as these can often confound findings and in line with recommendations from reviews in the field (Murray et al., 2017; Sami & Bhattacharyya, 2018). Furthermore, importantly the PwP + C and PwP – C patient groups are well matched across clinical parameters, including diagnosis functioning, symptomatology, days of hospitalisation and medication, indicating that any difference found cannot be accounted for by clinical differences. For shape analysis we used stringent criteria for multiple comparisons correction (TFCE, FWE < 0.5). FIRST undertakes analysis on a structure by structure basis, although it is arguable whether this would be any different if cluster correction was applied across the total number of subcortical surface vertices of all shapes.

Notwithstanding this we also note limitations to the study design. The study was cross‐sectional which precludes inference on the precise nature of causation between groups. We limited analysis to the subcortical deep grey matter structures and did not look at cortical thinning, cerebellar volumes, white matter tracts or ventricular enlargement, which may also be impacted by cannabis use. Furthermore the sample size was relatively modest to detect volumetric differences (Koenders et al., 2014) although comparable to previous such studies in shape (Smith et al., 2014; Smith et al., 2015; Solowij et al., 2013). For this reason, we did not try to disentangle gender‐specific effects, although there was no significant difference in gender between groups.

Taken together we have shown a neurobiological distinction between those in early psychosis with and without a history of cannabis use in the form of bilateral regional putamen enlargement. The putamen and thalamus may be key neurobiological substrates for compulsivity in the dual diagnosis group and further research should aim to delineate the behavioural and neurobiological correlates of these structures.