We tested two oculomotor endophenotypes of schizophrenia in patients with early psychosis and controls with and without a history of cannabis use. We provide evidence for patient and cannabis effects in SPEM velocity gain and a cannabis × patient interaction such that there is a large difference between EPNC vs EPC. In addition to the statistical analysis described we would also draw the readers’ attention to a qualitative appreciation of Fig. 2. As stimulus frequency increases, gain decreases and variance increases for all groups, as expected. The EPNC group (green bar) shows both impaired gain and increased variance compared to the other groups. In contrast, the EPC group (beige bar) behaves more in keeping with control groups. Intriguingly, we show a deficit in EPNC compared to EPC of large effect size. However, we do not find a patient or cannabis effect for AS error. Taken together, the EPNC group shows impairment in SPEM velocity gain whereas this is spared in the EPC group, but no such difference is shown in AS error rate. We also observed that patients made significantly fewer valid AS trials compared to controls, which may indicate generally poorer task performance.

These findings are of particular interest as the two patient groups were, with the exception of substance use measures, essentially indistinguishable clinically, and none of the covariates (Fagerstrom and AUDIT scores) were significantly associated with either oculomotor measure.

SPEM abnormalities are considered a trait marker unrelated to clinical state or progression of disease^20,23,25,26 and there is good evidence from familial and genetic approaches that impairment in SPEM indexes genetic risk of schizophrenia^28,43. One explanation of our findings, within the confines of a cross-sectional study, may be that there is less biological predisposition to psychosis for EPC compared to EPNC. This would, consistent with the decreased vulnerability hypothesis, imply that a history of cannabis use has an additive effect on psychosis liability and hence that some in this group (particularly those with the lowest predisposition to psychosis) would not have developed psychotic disorders without cannabis use, which pushed them over the psychotic threshold (Fig. 1).

An alternative explanation for our findings could be that over time chronic cannabis use may have normalised the SPEM abnormality associated with psychosis liability that may also have been present in EPC at onset of psychosis. A recent study of meditators and non-meditators showed that cultivated mindfulness is associated with fewer saccadic intrusions in SPEM. One could conceivably argue that cannabis has a similar effect for patients with psychosis through similar cognitive and attentional processes, thereby normalising SPEM⁴⁵. It is worth noting, however, that the Kumari and colleagues findings were related to saccadic intrusions during SPEM, finding no difference in SPEM velocity between meditators and non-meditators and, therefore, does not suggest that SPEM velocity, as indexed in our study is of itself is a measure of cultivated mindfulness. A mechanism where chronic cannabis use normalises SPEM would suggest eye movements indexes state rather than trait components. In a post hoc supplementary analysis we did not find direct evidence that SPEM shows a state component in our study (post hoc correlation of SPEM with PANSS scores; p > 0.15). However, we accept that it is not fully possible to disentangle these explanations within the confines of a cross-sectional study. A better design to test this would be to measure eye movements before and after onset of chronic cannabis use. We note that acute THC or cannabis use has been shown to decrease rather than increase SPEM velocity gain in the previous literature^33,34.

Our results are in line with evidence suggesting a better neuropsychological profile in cannabis-using patients at onset of psychosis compared to non-users⁴⁶. It is however difficult to definitively address this issue as cannabis use itself may impair neuropsychological performance, and hence results have been mixed⁴⁷. Once psychotic disorder has developed, use of cannabis indicates a poorer prognosis^8,48. After 5 years of disorder there is evidence of detrimental effects in structural MRI measures (hippocampal, subcortical atrophy and cortical thinning) in cannabis-using patients compared to non-using patients⁴⁹. Our results also newly identify a biomarker allowing a neurobiological distinction between these groups. Taken together, the concept of a cannabis using group with lower biological predisposition at psychosis onset but worse neurobiological indices and prognosis after long-term use indicates the utility of putatively considering cannabis-using psychosis as a distinct nosological entity. Longitudinal studies are needed to substantiate these conjectures.

It is of interest that we found differences in SPEM performance but not in relation to AS. Possible reasons for why this may be are necessarily speculative. First, although both SPEM and AS are candidate endophenotypes of schizophrenia, emerging evidence demonstrates that they are associated with differential genetic architecture²⁸. It is possible, therefore, that EPC and EPNC groups differ in expression of SPEM related genes but not AS genetic expression. Exactly which genes are implicated requires further delineation. Secondly, and non-mutually exclusively it is possible that certain neurobiological substrates of the visual processing pathway are implicated in this difference which involve SPEM but not AS, such as the magnocellular system and NMDA-related mechanisms^50,51. Specifically, the basic afferent pathway (retina–>lateral geniculate nucleus–>V1) is identical in both tasks, but differences emerge in the role of cortical regions (medial temporal area, medial superior temporal area, frontal eye fields (FEF) with modulation from the supplementary eye field and parietal cortical areas and a loop to caudate). Whilst AS involves prominent top-down control⁵², SPEM relies on target selection, motion processing, prediction and feedback processes⁵³. SPEM and AS show overlapping but differential patterns of activation although definitive neurobiological substrates underpinning this distinction remain to be replicated^54,55. Hence further work using functional MRI of eye movements within the scanner would be needed to delineate which specific areas are implicated in the observed differences between groups.

Particular strengths of this study include the use of robust and well-established biomarkers. A further strength of this study is that both patient groups were well matched across clinical parameters. This study also avoided some of the limitations of other studies comparing EPC and EPNC⁴⁹: we accounted for tobacco or alcohol use which are more common in cannabis users⁵⁶ and included a cannabis control group^49,57. Finally, this study is undertaken in a real-world clinical dataset and likely to be representative of the broader clinical population.

There were limitations of this study. Groups were moderately sized, although comparable to previous eye-tracking studies, and it is possible that the inability to detect group difference in error rates in the AS task was due to low power. The cross-sectional design means it is difficult to disentangle precise casual relationships. Further work would ideally be longitudinal, charting the transition to psychotic disorders until the illness is established, or including clinical high-risk groups. We covaried for alcohol and tobacco measures, and comorbid substance use disorders were not significantly different across groups. However, in an observational setting it is not possible to completely exclude the effect of other substances of abuse, such as nicotine. We should note as discussed above that a state component to the cannabis effects cannot be precluded. Finally, whilst this study provides important new data about the EPC group, where cannabis has been heavily used as the primary drug of abuse, it is not clear whether these results are generalisable to cases of poly-substance use or patients in whom cannabis use is of mild intensity.

Notwithstanding these limitations, taken together we provide evidence for group-level differences in SPEM mean velocity gain between patients with and without a history of cannabis use. This may indicate a difference in biological diathesis to psychosis between groups. Further work is required to delineate the neurobiological substrates for this distinction.